点击显示 收起
1 From the Sogn Center for Child and Adolescent Psychiatry, University Hospital (AMM), the Institute for Nutrition Research (MHC, SKB, HLW, and RB), and the Section of Medical Statistics, Faculty of Medicine (PL), University of Oslo.
2 Supported by the Research Council of Norway and Johan Throne Holst’s Foundation. 3 Address reprint requests to R Blomhoff, Institute for Nutrition Research, Faculty of Medicine, University of Oslo, PO Box 1046, Blindern, 0316 Oslo, Norway. E-mail: rune.blomhoff{at}basalmed.uio.no.
ABSTRACT
Background: It is well established that an excessive intake of retinol (vitamin A) is toxic; however, it has been > 25 y since the last extensive treatise of case reports on this subject.
Objective: The objectives were to identify and evaluate all individual cases of retinol toxicity published in the scientific literature that assessed the thresholds and symptoms induced by high intakes of retinol and to compare the toxicity of different physical forms of retinol preparations.
Design: We performed a meta-analysis of case reports on toxicity claimed to be induced by intakes of excessive amounts of dietary retinol (ie, retinol and retinyl esters in foods or supplements). Using free text and MESH (medical subheading) strategies in PubMed, we identified 248 articles in the scientific literature. From these initial articles we identified other relevant citations. The final database consisted of 259 cases in which individual data on dose, sex, age, time of exposure, and symptoms are reported.
Results: Chronic hypervitaminosis A is induced after daily doses of 2 mg retinol/kg in oil-based preparations for many months or years. In contrast, doses as low as 0.2 mg retinol · kg-1 · d-1 in water-miscible, emulsified, and solid preparations for only a few weeks caused chronic hypervitaminosis A. Thus, water-miscible, emulsified, and solid preparations of retinol are 10 times as toxic as are oil-based retinol preparations. The safe upper single dose of retinol in oil or liver seems to be 4–6 mg/kg body wt. These thresholds do not vary considerably with age.
Conclusions: The results of the present study indicate that the physical form of retinol supplements is a major determinant of toxicity. The use of water-miscible, emulsified, and solid preparations of retinol should therefore be carefully considered before being used in supplements and fortifications.
Key Words: Vitamin A toxicity • hypervitaminosis A • fortification • retinol • retinyl esters • water-miscible retinol • emulsified retinol • supplements
INTRODUCTION
A large intake of dietary retinol (ie, vitamin A) is toxic. Such toxicity is classified as acute, chronic, or teratogenic (1–3). Acute hypervitaminosis A is due to a single or a limited number of large doses taken during a short period, whereas chronic hypervitaminosis A is caused by moderately high doses taken frequently, usually daily, over a span of months or years. Teratogenicity is caused by the ingestion of moderate-to-high doses during the first trimester of pregnancy in humans (1–3).
Chronic hypervitaminosis A in humans has been reported after recurrent intakes of retinol in amounts =" BORDER="0"> 10 times the recommended dietary allowance (RDA), ie, =" BORDER="0"> 10 and =" BORDER="0"> 3.75 mg or more for adults and infants, respectively (1–3). Occasionally, however, reports have suggested that intakes marginally above the RDA also are associated with chronic hypervitaminosis A and teratogenicity. For example, a higher risk of embryonic malformations was observed among infants born to women who consumed > 4.5 mg retinol/d (median dose: 6.5 mg) from food and supplements than in infants whose mothers consumed 1.5 mg/d (4). Also, daily intakes by adults of as low as 1.5–3.0 mg retinol are associated with a reduction in bone mineral density and an increased risk of hip fractures (5, 6). If these associations are the results of a causal relation, a relatively large proportion of the population in many countries has an increased the risk of osteoporosis and teratogenicity because of excessive intakes of dietary retinol. Because it has been > 25 y since the last extensive treatise of case reports about retinol toxicity in the scientific literature (7) and because no meta-analysis of this issue has been performed, we evaluated statistically all individual cases of retinol toxicity published in the scientific literature.
SUBJECTS AND METHODS
Using various free text and MESH (medical subheading) strategies at PubMed, we identified articles reporting toxicity due to ingestion of retinol (ie, free retinol or retinyl esters). We excluded all articles dealing with retinoic acid or other retinoid medications. From these initial articles we identified other articles not included in PubMed (National Library of Medicine, Bethesda, MD), ie, articles published before 1965. We identified a total of 248 relevant articles. (See E-table with individual data on age, sex, nationality, weight, dose, condition, exposure time, form of retinol preparation, intake of other supplements, symptoms, and clinical-chemical analyses on the Internet: http://www.blomhoff.no.) Cases with toxicity due to high retinol intakes in clinical trials were not included when data on individual characteristics were not available. All articles were examined by =" BORDER="0"> 2 scientists. When symptoms of toxicity appeared within 24 h of intake of one or a few doses of retinol, the intoxication was defined as acute hypervitaminosis A.
The database consists of 291 cases reported between 1944 and 2000. Of these, 27 were excluded from further analysis because of limitations in the data; in most cases, the dose was not reported. The final database included 259 cases of hypervitaminosis A and 5 cases of teratogenicity.
To calculate doses per kilogram body weight, we used published weight curves for subjects aged 2–14 y (8). For subjects aged 0–2 y, the dose per kg was calculated by using the weight curve derived from those subjects in the same age group that reported body weights (n = 44). For subjects aged =" BORDER="0"> 15 y, representative weight curves were used (A Tverdal, Statens Helseundersøkelser, Norway, personal communication, 2002).
All reported P values are two-sided, and the significance criterion of P = 0.05 was used. We compared groups by estimating differences, P values, CIs, and test statistics, based on the Mann-Whitney U and Kruskal-Wallis tests. P values were Bonferroni adjusted for multiple comparisons. Statistical tests were performed with SYSTAT, version 9 (SYSTAT Software Inc, Richmond, CA).
RESULTS
The 259 cases of hypervitaminosis A originated from 23 different countries, but most of the cases were from the United States (n = 105), France (n = 56), and Spain (n = 24) (see E-table). Because the frequency of symptoms varied in children and adults, we separated the cases into age groups. In the age groups 0–2, 3–16, and =" BORDER="0"> 17 y, 100, 41, and 118 acute plus chronic cases of hypervitaminosis A were identified, respectively. The sex ratio (male/female) in the same age groups was 1.11, 1.40, and 0.39, respectively. Of those reporting the source of retinol, 4, 24, 36, and 32 cases came from ingestion of retinol in liver, oil, emulsified and water-miscible solutions, and solid tablets, respectively.
There were 55 cases with acute and 204 cases with chronic hypervitaminosis A. Of the 259 cases registered, we identified length of exposure in 243 cases. A graphic representation of dose (mg retinol · kg-1 · d-1) versus time of exposure in these 243 cases is presented in Figure 1. The group with chronic hypervitaminosis A consumed less retinol than did the group with acute hypervitaminosis A, but there was also a considerable overlap in dose (Table 1). The difference in median values between the groups (all ages combined) was 12.5 mg/kg (P < 0.0005; 95% CI: 10.6, 14.4 mg/kg). For the acute cases, all age groups appeared to be equally sensitive to retinol intoxication. For the chronic cases, however, the age group 0–2 y reported higher median daily doses per kg body wt than did those aged =" BORDER="0"> 3 y (P < 0.0005).
FIGURE 1.. Graphic representation of the 243 acute and chronic cases of hypervitaminosis A in the meta-analysis database for which the duration of intake was identified.
View this table:
TABLE 1. Cases of acute and chronic hypervitaminosis A
Water-miscible, emulsified, solid, and oil-based forms of vitamin A supplements
For 92 of the chronic cases, the physical form (ie, water-miscible, emulsified, solid, or oil-based preparations) of the vitamin preparation was given. We combined the data for the water-miscible and emulsified preparations in the statistical analysis because these forms often are interchangeable, although they are not identical. There was a significant difference reported between the doses of the 3 forms. Interestingly, the doses of the solid and water-miscible + emulsified retinol preparations were significantly lower than were the doses of the oil-based retinol preparations. The difference in median values between the solid and oil-based preparations was 3.8 mg/kg (95% CI: 2.1, 4.9 mg/kg; P < 0.0015) and that between the water-miscible + emulsified forms and the oil-based preparations was 3.5 mg/kg (95% CI: 1.2, 4.4 mg/kg; P = 0.006). The dose of solid preparations was also significantly lower than the dose of water-miscible + emulsified preparations; the difference in median values was 0.4 mg/kg (95% CI: 0.1, 0.7 mg/kg; P = 0.012). A difference in dose in these groups was also observed when the data were plotted as dose against duration of intake before the appearance of symptoms (Figure 2). For oil-based retinol supplements, daily doses of > 2 mg retinol/kg for weeks or months seemed to be needed for the induction of chronic hypervitaminosis A. When retinol was administered in a water-miscible or emulsified form, however, the threshold dose was reduced dramatically, and toxicity was evident after an intake of 0.2 mg retinol · kg-1 · d-1 for 1–4 wk.
FIGURE 2.. Duration of vitamin A intake before the first symptoms of toxicity appeared relative to the dose of retinol in subjects with chronic cases of hypervitaminosis A who had taken water-miscible or emulsified (), oil-based (), or solid (•) preparations of vitamin A. n = 54.
Vitamin D reduces toxicity
Retinol was taken together with vitamin D or alone in 39 and 42 chronic cases, respectively (Table 1). Vitamin D appears to protect against retinol toxicity because the median dose was significantly higher when the vitamins were combined (0.7 mg/kg; P = 0.020; 95% CI: 0.082, 1.56 mg/kg). The effect of vitamin D was independent of the physical form of the supplement. Of the subjects who took retinol together with vitamin D, 13 took oil-based preparations, 8 took water-miscible or emulsified preparations, and 18 took preparations for which the form was unknown. For subjects who did not take vitamin D preparations in combination with retinol, 1 subject took an oil-based preparation, 5 subjects took water-miscible or emulsified preparations, 2 subjects took solid tablets, and 34 subjects took preparations for which the form was unknown.
Symptoms reported in acute hypervitaminosis A
Fifty of the subjects with acute hypervitaminosis A were in the age group 0–2 y (Table 1). Most of these subjects experienced symptoms of the nervous, vision, and gastrointestinal systems. General symptoms of a deteriorating state of health (eg, fever, loss of appetite, and fatigue) were reported by 36% of the group (Table 2). For the 5 subjects aged > 2 y, symptoms of the skin and hair, of vision, and of the nervous and gastrointestinal systems were reported most frequently (data not shown). Although vomiting can also be a result of increased central nervous system pressure (or pseudotumor cerebri), we categorized it as a gastrointestinal symptom in the present study.
View this table:
TABLE 2. Percentage occurrence of symptoms registered in cases with acute or chronic hypervitaminosis A in different age groups1
Symptoms reported in chronic hypervitaminosis A
For the age groups 0–2 and 3–16 y, symptoms of the skin; symptoms of the vision, nervous, gastrointestinal, and musculoskeletal systems; and general symptoms of a deteriorating state of health were reported in > 50% of the chronic cases (Table 2). Adult subjects experienced mostly symptoms of the gastrointestinal system and skin and hair and symptoms of a deteriorating state of health. For specific symptoms, see Table 2.
Symptoms and intake of different forms of vitamin A supplements
In the age group 0–2 y, we identified 35 subjects who had taken either oil-based or water-miscible or emulsified retinol preparations. For the subjects with intakes of oil-based preparations, the symptom groups musculoskeletal system, skin and hair, and nervous system and vision and symptoms of a deteriorating state of health were reported by most of the subjects (Table 3). However, only symptoms of the nervous and vision systems and symptoms of a deteriorating state of health were reported by most of the subjects who consumed water-miscible or emulsified retinol. Strikingly, only 50% and 28% of the subjects with intakes of water-miscible or emulsified retinol reported symptoms of the skin and hair and of the musculoskeletal system, respectively. Of the specific symptoms, bulging of the fontanels was linked to the intake of retinol in water-miscible or emulsified forms, whereas hyperostosis, alopecia, and skeletal pain were more often linked to the intake of retinol in oil. In general, subjects who took oil-based retinol supplements reported more symptoms than did those who took water-miscible or emulsified forms.
View this table:
TABLE 3. Percentage occurrence of symptoms registered in chronic cases in which the forms of retinol preparations were identified1
For the age group =" BORDER="0"> 17 y, 48 subjects had taken either water-miscible + emulsified or solid retinol preparations. Subjects that had taken water-miscible + emulsified retinol more frequently experienced symptoms of the nervous system, of vision, and of the skin and hair, whereas subjects that had taken solid retinol more frequently experienced symptoms of the gastrointestinal system and symptoms of a deteriorating state of health (Table 3).
Serum markers of toxicity
We identified 19 different serum markers. Elevated concentrations (given as a percentage of the subjects for whom these markers were measured) were reported for serum retinol (ie, retinol and retinyl esters) (89%, n = 90), serum -glutamyltranspeptidase (83%, n = 6), serum lipids (80%, n = 5), triacylglycerol (75%, n = 8), alkaline phosphatase (67%, n = 55), prothrombin time (53%, n = 32), cholesterol (50%, n = 14), aspartate aminotransferase (49%, n = 51), bilirubin (48%, n = 27), and calcium (46%, n = 52).
Teratogenic cases
Five cases of presumed or suspected teratogenicity due to chronic retinol toxicity were identified (see E-table). The mothers ingested 0.12–0.74 mg retinol · kg-1 · d-1 (7.2–44.4 mg/d for a woman weighing 60 kg) in the first trimester of pregnancy. In 4 of the chronic cases, renal anomalies were reported in the infants. In the fifth case, the only anomaly reported was a cleft anterior segment of one eye. The 2 mothers with the highest intakes of retinol (0.73 and 0.74 mg retinol · kg-1 · d-1) delivered infants with more massive anomalies, including malformations of the central nervous system.
DISCUSSION
Threshold dose of chronic toxicity and intake of different forms of retinol supplements
The present study suggests that mild symptoms of chronic hypervitaminosis A may be induced if daily doses of 2 mg oil-based retinol/kg is ingested for > 1 y. If retinol is administered in a water-miscible, emulsified, or solid form, however, the threshold dose seems to be reduced dramatically and toxicity appears earlier. The threshold dose observed in our meta-analysis may be distorted because of selection bias, because, obviously, not every case of retinol toxicity is expected to be published. On the other hand, toxicity due to other unknown or uncontrolled factors may also become part of the scientific literature as sole cases of retinol toxicity. There is no reason to believe, however, that publication bias is different for studies in which toxicity is reportedly due to the form of retinol supplement taken. Thus, our data strongly suggest that water-miscible, emulsified, and solid forms of retinol supplements are significantly more toxic than are oil-based forms. Our results confirm and extend the results of an earlier small-scale study by Korner and Vollm (7) 25 y ago. The finding of the greater toxicity of water-miscible and emulsified forms of retinol also agrees with the results of studies in experimental animals, infants, children, and adults, in which it was observed that water-miscible and emulsified retinol result in higher plasma peak values, higher liver concentrations, and lower fecal losses than do oil-based retinol preparations and retinol in liver (9–14).
Our suggested threshold doses of 2.0 and 0.2 mg retinol/kg in oil and in water-miscible + emulsified forms, respectively, should be considered with care. These doses, however, agree with threshold doses reported in human intervention studies. Clinical studies of secondary cancer prevention indicate that daily doses of 90 mg retinol in oil-based preparations in adults (1–1.5 mg/kg) are well tolerated for many months or years (15, 16). The adverse clinical side effects reported after 1–2 y of treatment were mild dermatologic symptoms in 40–55% of the intervention group. However, a similar dose of retinol (ie, 90 mg to adults, or 1–1.5 mg/kg) in a water-miscible + emulsified form given daily for 12 mo in an Italian study resulted in earlier and more pronounced side effects (17), such as increases in the concentrations of -glutamyltranspeptidase and serum triacylglycerol and dermatologic symptoms (18, 19).
The RDA for retinol is generally set between 0.8 and 1.0 mg/d (20), which is equivalent to 0.011–0.014 mg/kg for an adult weighing 70 kg. The lowest dose reported to cause acute hypervitaminosis A in adults, regardless of the physical form of the preparation, is 0.02 mg/kg. For children with chronic hypervitaminosis A, the lowest observed toxic dose is in the range of the RDA for adults. The upper limit for men aged =" BORDER="0"> 19 y and for women aged =" BORDER="0"> 51 y is 3 mg preformed retinol/d (20), which is equivalent to 0.04 mg retinol · kg-1 · d-1 for a person weighing 70 kg. Thus, a few cases of toxicity from doses near the RDA and below the upper limit have been reported (Table 1). These cases most likely represent persons who suffer from retinol intolerance, ie, a relatively rare condition that seems to be genetic and mainly affects males (1–3). Alternatively, these cases may arise from combined toxicity because alcohol ingestion, low protein intakes, viral hepatitis, some environmental pollutants and drugs, and diseases of the liver and kidney are known to enhance retinol toxicity (1–3).
Vitamin D is associated with reduced vitamin A toxicity
Interestingly, we observed that vitamin D appears to protect against retinol toxicity because the dose was significantly higher when vitamin A and vitamin D were combined. Although it has not been suggested that vitamin D protects against hypervitaminosis A, an antagonistic relation between retinol and vitamin D has been observed in several experimental systems (21–23).
Threshold dose and symptoms of acute hypervitaminosis A
Acute hypervitaminosis A results mainly in symptoms of the nervous, vision, and gastrointestinal systems. Only 5 cases of acute hypervitaminosis A have been reported among persons aged =" BORDER="0"> 3 y. In these cases the dose ranged from 6.3 to 37.6 mg/kg, which is equivalent to 440–2500 mg for an adult. This finding agrees with that in Arctic populations and explorers in whom intakes of 250–1000 mg retinol from the liver of polar bear and bearded seal induced acute toxicity (24). In comparison, an intake of 150 g liver from domestic animals, which is frequently eaten in several countries without adverse side effects, contributes to an intake of 25–75 mg retinol. Also, to test fat malabsorption, administration of 60–150 mg retinol in oil has been regarded as a safe test, although it is rarely performed (25–27). Thus, the safe upper single dose of retinol in oil or liver seems to be 250–440 mg retinol for adults (ie, 4–6 mg/kg).
The lowest dose reported for subjects aged 0–2 y was 3.5 mg/kg. Information about safe doses for infants and small children is also obtained from intervention programs in which large doses of retinol are administered in immunization programs to combat vitamin A deficiency. In such programs, side effects are rare in infants, although there are a few reports of toxic effects in the first 6 mo of life. Two double-blind, randomized, placebo-controlled trials in Bangladesh reported higher incidences of bulging of the anterior fontanel in children treated with water-miscible or emulsified retinol (7.5–15 mg retinol) than in untreated control subjects (28, 29). The effects were transient and not accompanied by any neurologic signs. Furthermore, a randomized, double-blind trial in Nepal (30) in neonates (< 1 mo of age) and infants (1–6 mo of age) who received a large oral dose of 15–30 mg retinol in oil, observed that neonates showed no excess risk of adverse side effects after they ingested 15 mg retinol. Compared with control subjects, the older infants who ingested 30 mg retinol had a 1.6% excess rate of vomiting and a 0.5% excess rate in the occurrence of bulging of the fontanel. Similar findings were also observed in studies in Indonesia, Guinea-Bissau, Ghana, India, and Peru (31–33). Thus, in accordance with our meta-analysis, the safe dose of retinol in oil for infants and small children is 3–3.5 mg/kg, whereas water-miscible and emulsified forms of retinol have a lower threshold.
Hypervitaminosis A, osteoporosis, and teratogenicity
In subjects aged 0–2 y, decalcification was reported by 11% of the children who received water-miscible or emulsified retinol, whereas none of the subjects who received retinol in oil reported this symptom. These data may be related to the results of some recent epidemiologic studies that link retinol intake to the risk of osteoporosis. Some (5, 6, 34–36) but not all (37–41) studies found that intakes of total vitamin A (ie, sum of provitamin A carotenoids and preformed retinol) or preformed retinol are associated with reductions in bone mass density and with increases in the risk of hip fractures.
Epidemiologic studies that reported an association between retinol intake and osteoporosis (5, 6) suggest that a threshold dose as low as 1.5 mg retinol/d (or 0.02 mg · kg-1 · d-1 in adults). For teratogenicity, the suggested threshold dose is 3 mg retinol/d (4). It would be interesting to know whether the physical form of retinol supplements influences the risk of osteoporosis and teratogenicity in a manner similar to that described in the present report. If this is indeed the case, a substantial proportion of the population may have an increased risk of osteoporosis and teratogenicity if they consume supplements and fortified foods containing retinol in a water-miscible, emulsified or dry physical form.
Conclusion
The present meta-analysis shows that it is important to take into account the physical form of retinol supplements when hypervitaminosis A is considered: water-miscible, emulsified, and dry preparations of retinol seem to be =" BORDER="0"> 10 times as toxic as oil-based preparations of retinol or dietary retinol in liver. Chronic hypervitaminosis seems to be induced after daily doses of 2 mg retinol/kg in oil-based preparations for many months or years. The safe upper single dose of retinol in oil or liver seems to be 4–6 mg/kg body wt. These thresholds do not vary considerably with age. Interestingly, vitamin D appears to protect against vitamin A toxicity.
ACKNOWLEDGMENTS
RB drafted the protocol and manuscript; AMM, MHC, HLW, and SKB performed the literature searches and extracted the data; SKB developed the database; PL and MHC performed the statistical analysis; MHC had a major role in the preparation of the manuscript; and all authors helped write the manuscript. RB is the guarantor of this article. RB received consultant fees from Roche, Basel, and Vitas, Oslo. AMM, MHC, HLW, SKB, and PL had no conflict of interest to report.
REFERENCES