Dietary Retinol—A Double-edged Sword

In this issue of THE JOURNAL, Feskanich and colleagues¹ report that women enrolled in the Nurses' Health Study (NHS) who had the highest intake of vitamin A and, specifically, retinol had higher rates of nontraumatic hip fracture than women with the lowest intake. This association raises important questions on the safety of long-term ingestion of dietary retinol. But should these observations alter recommendations for retinol or provitamin A intake from foods, fortified foods, and supplements? To answer, the sources and functions of dietary vitamin A must be considered.

The term vitamin A refers to a family of essential, fat-soluble dietary compounds required for vision, growth, reproduction, cell proliferation, cell differentiation, and the integrity of the immune system.² The most potent vitamin A compound, all-trans retinol, is capable of reversing signs and symptoms of vitamin A deficiency. Although all-trans retinol has been retained as the standard against which vitamin A activity is measured, it is now well established that all-trans retinoic acid is responsible for nearly all of the biological effects of vitamin A except those of phototransduction.³ Although the 2-step process that creates all-trans retinoic acid is well characterized, how humans regulate the irreversible step of oxidation into a carboxylic acid is poorly understood.

All-trans retinoic acid exerts its biological effects when bound to any one of several members of the retinoic acid receptor (RAR)/retinoid X receptor (RXR) nuclear transcription factor family. Found in nearly every cell, RAR/RXR transcription factors can increase or decrease gene expression by binding to specific DNA response elements. The ability of retinoic acid to regulate gene products in spatial and temporal patterns accounts for the programmed cell growth, differentiation, and apoptosis essential for normal embryonic development, cellular differentiation, and immunity. Retinoic acid is a powerful mediator of cell function. Too much or too little retinoic acid creates a disruption of these tightly regulated processes. It is therefore not surprising that dietary retinol deserves cautious respect.⁴

The human diet is biased against excessive production of retinoic acid because of the paucity of sources for dietary retinol. Retinol esters are naturally present in the fat of only a few foods; the most concentrated sources are liver, fish liver oils, eggs, and whole-milk dairy products. The human body has adapted to major day-to-day variation in retinol intake by creating tightly regulated storage pools for dietary retinol in the plasma and liver.

The most important source of dietary retinol is animal liver.⁵ Despite its infrequent consumption, liver contributes as much dietary retinol to the US diet as that contributed by whole milk and eggs combined. A 3-oz (84-g) portion of liver contains 100 times the retinol present in 1 glass of whole milk, 116 times that in 1 whole egg, and 200 times that in a pat of butter.⁶ Vitamin A intake variability is further affected by the 3-fold natural seasonal variation in the vitamin A content of milk.⁷

Retinol can also be found in fortified foods. Since milk and butter are important sources of retinol, the US Food and Drug Administration (FDA) defined the identity of low-fat and nonfat milk as well as all margarines to contain 150 µg/c (624 µg/L) and 55 µg per pat of all-trans retinol, respectively, to preserve the nutritional quality of these dietary substitutes.⁸ An additional source of dietary retinol that has not been subjected to FDA regulation is the fortification of breakfast cereals and meal replacement beverages. This latter source of dietary retinol is increasing in importance and provides for women an amount of retinol equivalent to that provided by fortified margarine and low-fat dairy products combined.⁹ Approximately two thirds of the US recommended daily intake of vitamin A is met by direct intake of dietary retinol.¹⁰

The remaining third of vitamin A in the US diet is obtained from consuming provitamin A. Carotenoids can serve as a source of provitamin A if they contain a β-ionone ring. Approximately 10% of dietary carotenoids contain at least 1 β-ionone ring, and beta carotene contains 2. The Institute of Medicine's recently released recommended daily intake guideline¹⁰ has reduced by half the amount of provitamin A carotenoids expected to be converted into retinol. Renamed "retinol activity equivalents," 12 µg of beta carotene or 24 µg of other provitamin A carotenoids are now considered to be equivalent to 1 µg of all-trans retinol. The poor conversion rate of carotenoids into vitamin A is due to the interference of vegetable matrix with absorption and from the presence of chimeric versions for each carotenoid, with every deviation from the all-trans structure leading to a diminution of retinol activity. Despite this low conversion rate, fruits and vegetables remain an important source of vitamin A in the United States; the most common sources are carrots, tomatoes, vegetable soups, greens, cantaloupe, and spinach. The bias toward low intake in the average human diet makes vitamin A deficiency the third most common nutrient deficiency in the world.³ It is particularly prevalent in countries where a combination of low meat and dairy intake, reduced variety of vegetable intake, and endemic roundworm infection compromise the absorbable vitamin A content of the diet.

In contrast with the usual low intake of vitamin A, Feskanich et al¹ found that a substantial proportion of women in their study had relatively high vitamin A intake and that high intake was associated with increased risk of hip fracture. Several aspects of this investigation bear review.

First, given the infrequent consumption of foods containing retinol and provitamin A, a food frequency questionnaire (FFQ) provides a different and perhaps superior method for estimating a given individual's true intake of vitamin A intake.¹¹ Because of the wide fluctuations in daily vitamin A intake, accurate classification of individual intake requires averaging intakes over a 1.5- to 2-month period.¹² The current study used an FFQ that queries intake of key foods containing vitamin A and averaged intake from up to 5 FFQs to minimize measurement error. Although any dietary assessment method is subject to errors of self-report, the FFQ more completely reviews the food sources of dietary retinol and provitamin A. The similarity between the estimated 1994 intake of dietary retinol from foods with the intake estimated from the Third National Health and Nutrition Examination Survey (NHANES III) 24-hour dietary recall¹⁰ adds further support that the specific estimates of vitamin A intake calculated in this investigation are reasonable.

However, the dietary practices of the NHS cohort differ from the dietary practices of the US population at large. The top quintile of average retinol intake from food (Table 2) is at the 90th percentile of similarly aged women in NHANES III and the bottom quintile corresponds to the 50th percentile of NHANES III women. Since more than half of the NHS cohort were taking multivitamins that contained retinol, total intakes were even higher. The wisdom of daily supplemental retinol in older persons has been questioned because of the finding that older but not younger supplement users have much higher levels of plasma retinyl esters.¹³ Whether higher plasma retinyl ester content predisposes elderly supplement users to more subtle signs of vitamin A toxicity, as suggested in the current investigation, has yet to be proven. Among the NHS cohort, beta carotene intake was also higher than that reported in NHANES III, but this finding is consistent with the observation from the US Department of Agriculture's finding that women of higher socioeconomic status consume more fruits and vegetables.¹⁴

A second issue is that despite the relative homogeneity of the study cohort, which may more easily permit detection of an untoward effect of dietary retinol, some confounding still could occur. For example, women with higher vitamin A intake consumed less alcohol, and moderate alcohol consumption has been associated with 10% to 12% higher bone mineral density.¹⁵ In a study of nearly 10 000 postmenopausal women, moderate drinking was significantly associated by univariate analysis with a 30% lower risk for hip fracture; the benefits diminished to a nonsignificant 20% reduction after adjustment for the generally healthier status of these moderate drinkers.¹⁶ In addition, alcohol is known to exacerbate retinol deficiency and retinol toxicity.¹⁷ The authors adjusted their findings to account for differences in alcohol intake, but whether they considered the reported J-shaped association between alcohol intake and fracture is unclear.¹⁸ What is most surprising is that the remaining potential confounders—calcium, vitamin D, and vitamin K—all would have contributed to finding a protective effect of vitamin A, not the untoward effect by Feskanich et al.¹ The association of higher risk of hip fracture only in postmenopausal women not taking estrogens adds further intrigue to the observation. Estrogens are known to block several steps in osteoclast function, including differentiation, activation, and programmed cell death,¹⁹ supporting the possibility that estrogen could oppose the type of effects expected of retinoic acid.

The new recommendations for vitamin A intake from the Institute of Medicine are 800 µg/d for men and 700 µg/d for women, with a safe upper limit of 3000 µg/d of retinol.¹⁰ A safe upper limit for provitamin A was not defined because no cases of vitamin A toxicity attributable to excessive provitamin ingestion have been reported. The findings of Feskanich et al provide further support for restricting a safe upper limit to dietary retinol alone and question whether the women represented by this study cohort—white US women of high socioeconomic status—should avoid supplements containing retinol if their diet is rich in low saturated fat, low-trans margarine, low-fat dairy products, and fruits and vegetables. Other populations with diets less rich in vitamin A will require further study to avoid confronting the other, even more hazardous side of the sword—vitamin A deficiency. In the meantime, this study serves as a reminder that vitamins are potent, essential nutrients which have effects that can precipitate harm as well as provide benefit. The optimal source is from the foods in our diets, not the dietary supplements often taken to simplify our complex world.