How Strong Is the Evidence That Nutritional Supplements Slow the Progression of Retinitis Pigmentosa?Nutritional Supplements and Retinitis Pigmentosa

In this issue of the Archives, Berson and colleagues¹ report the results of a well-designed and well-executed clinical trial that extends their studies of the effects of nutritional supplements on the rate of progression of retinitis pigmentosa (RP). Their first such study, published in the Archives in 1993, investigated the effects of consuming 15 000 IU/d of vitamin A and 400 IU/d of vitamin E on the rate of decline in the amplitude of the 30-Hz flicker electroretinogram (ERG). Their second trial, published in the Archives in 2004, tested the effect of taking 1200 mg/d of docosahexaenoic acid (DHA) combined with a high dose of vitamin A on the rate of visual field loss in RP. The third trial, published in this issue of the Archives, examined the supplemental effects of 12 mg/d of lutein combined with the high dose of vitamin A and recommended high dietary intake of DHA on the rate of RP visual field loss.

None of these studies produced simple, clear-cut results. In the 1993 study, patients were randomized to 1 of 4 groups: vitamin A, vitamin E, vitamin A plus vitamin E, or placebo. When the combined results of the A and A+E groups were compared with the combined results of the E and control groups, the rate of log_e ERG amplitude loss over 4 years was reported to be significantly slower for the combined A and A+E groups than it was for the combined E and control groups (P < .01).² However, a few months later, members of the study's Data and Safety Monitoring Committee reported in letters to the Archives that there was no difference between the vitamin A group and the control group or between A+E group and the control group; they argued that much of the originally reported significant difference was a consequence of pooling the data and could be attributed to early and consistently large differences between the vitamin E group and all of the other groups.^{2 - 3}

The primary outcome measure in the 2004 study was the annual rate of loss in the total point score on the Humphrey Field Analyzer (HFA) 30-2 static visual field test. The investigators reported that there was no difference in rates of loss over 4 years between the treatment and control groups (P = .88).⁵ There also was no difference between treatment and control groups for any of the secondary outcome measures (rates of loss in 30/60-2 HFA total point score, 30 Hz ERG log_e amplitude, and logMAR visual acuity). However, in a follow-up article published in the same issue of Archives, the investigators reported the results of a subgroup analysis and concluded that of the patients who did not take vitamin A prior to the study, the DHA supplement slowed the rate of HFA total point score loss (P = .01) and logERG amplitude loss (P = .03) in years 1 and 2, but not in years 3 and 4, after the start of treatment.⁶

Based on the conclusions of their analyses in the first study, the investigators recommended that patients with RP consume 15 000 IU/d of vitamin A palmitate to slow the progression of retinal degeneration. Although the investigators refrained from making a formal clinical practice recommendation, they did conclude from their analyses in the second study that a dietary supplement of DHA facilitates and hastens the therapeutic benefit of high daily doses of vitamin A in the first 2 years of treatment.

In the lutein trial in this issue of the Archives, patients again were followed up annually for 4 years. The primary outcome measure was the rate of decline in the total point score for the HFA 30-2 program using the largest stimulus (size V). Secondary outcomes were the rates of decline in the total point scores for the 60-4 program and for the 30-2 and 60-4 combined programs. Secondary outcome measures also included the rates of decline in the log_e amplitude of the 30-Hz ERG and logMAR visual acuity measured with the Early Treatment Diabetic Retinopathy Study chart.

The investigators reported no difference between groups in the rate of loss for the primary outcome measure (P = .66). Similarly, no differences were observed in secondary outcome measures between groups for rates of loss of the HFA 30-2 plus 60-4 total point score (P = .24), log_e ERG amplitude (P = .59), or logMAR visual acuity (P = .80). However, the investigators reported a significant effect of treatment on the rate of loss for the HFA 60-4 total point score (P = .05) and conclude that patients with RP would benefit from taking 12 mg of lutein per day in addition to the previously recommended 15 000 IU/d of vitamin A and weekly meals of oily fish to slow RP progression, particularly visual field loss in the midperipheral ring tested by the HFA 60-4 program.¹

This series of trials raises 2 important issues: (1) the risk of misinterpreting outcomes because of assumptions built into the data analysis, and (2) the question of how much weight should be placed on clinical practice recommendations that are based on secondary analyses. With respect to the first issue, comparing rates of loss in functional measures implies a particular dynamical model of the disease progression. Retinitis pigmentosa progresses slowly and, depending on the choice of measure, the loss might appear to follow a linear decline over time or might appear nonlinear; for example, it may follow an exponential decay. If the loss is nonlinear but the analysis assumes it is linear over the interval of interest, then the estimated rate of loss will depend on the stage of progression at the time the measurement is made. This dependence can lead to misinterpretations of outcomes, particularly if the groups are not well balanced with respect to stage of progression.

In the lutein study by Berson et al,¹ the primary and 2 of the secondary outcome measures were based on HFA total point scores, which are summed sensitivities in decibel units over all tested points in the visual field that had measurable contrast thresholds. The implicit assumption of the analysis used by the investigators is that over the natural course of RP, the decline of HFA total point score is linear with time. If this assumption about RP dynamics is correct, then the estimated rate of decline will be independent of the total point score at baseline. Ideally, this assumption should be tested using the raw data; however, as shown in our Figure, the trend can be demonstrated using the summary data published in the study's Tables 3 and 4. Instead of falling on a horizontal line, as predicted by the assumption of a linear decline over time, the estimated rate of decline for both the treatment and control groups is proportional to the HFA total point score at baseline. This straight-line relationship between the rate of decline and the baseline total point score means that the decline in total point score with time follows an exponential decay. This result suggests that if the investigators were to analyze the rate of decline in the logarithm of the total point score as they did with ERG amplitude (N.B., the logarithm of an exponential decay is linear), they might draw a different conclusion about the effectiveness of lutein treatment. The result in the Figure is not particularly surprising because numerous longitudinal RP natural-history studies have shown that log_e ERG amplitude^{7 - 12} and log_e visual field area^{7 - 18} decline linearly with time.

The second issue we want to raise is much thornier. Although the studies discussed in this editorial have produced a number of scientifically interesting observations that warrant further investigation, we are not convinced by the evidence that any of the tested nutritional supplements influence the rate of RP progression. Analogous to Pascal's wager on the existence of a deity, one could argue that because there is no other viable treatment for RP and the conclusions drawn by the investigators are reasonable, why not cover our bets and recommend these treatments to patients with RP in case the investigators are right? With the exception of potential mercury toxicity, few people would argue that eating oily fish twice per week poses a health risk, and with the exception of potential Escherichia coli contamination, few people would argue that a daily diet of green leafy vegetables high in lutein would pose a health risk. However, the DHA and lutein trials were designed on the premise that these supplements could only add to the already accepted beneficial treatment effects of high daily doses of vitamin A.

The members of the 1993 study's Data and Safety Monitoring Committee who wrote letters to the editor convincingly argued that without the vitamin E group to provide contrast, there is no case to be made that vitamin A therapy is any better than a placebo in slowing the rate of progression of RP. The investigators made strong arguments based on secondary analyses of subgroup data that kept the idea alive but the secondary analyses did not prove by current standards of evidence that vitamin A therapy is effective. There is ample evidence that vitamin A in high doses can be toxic and can cause birth defects. That is why pregnant women and patients with compromised liver function were excluded from the 3 trials and why the investigators recommend that patients who receive the vitamin A therapy have annual liver function tests and monitor their bone health. So in the case of vitamin A therapy, there is a potential risk.

Unlike the sometimes casual “dialogue with nature” that can characterize a thread of scientific investigation and discovery, the clinical trial is a special class of study. It is highly formal in its design, structure, and execution and is motivated by a clinical question that, when resolved, will define a new standard of practice. The primary outcome measure, which is agreed on in advance and around which the study is designed, renders the verdict. The secondary measures and ad hoc analyses provide the supporting science. They help explain and embellish our understanding of the primary effect, can open new directions of investigation, and sometimes lead to serendipitous discoveries. The 3 trials reviewed here are examples of well-executed studies and they present a number of provocative results. However, none of the studies convincingly proved that the treatments put on trial are effective in slowing the rate of progression of RP and, therefore, do not warrant mandating a change in how patients with RP are treated.

Correspondence: Dr Massof, Lions Vision Center, 550 N Broadway, Sixth Floor, Baltimore, MD 21205 (rmassof@lions.med.jhu.edu).

Financial Disclosure: None reported.