Contents - Previous - Next

This is the old United Nations University website. Visit the new site at

Natural food sources of vitamin A and provitamin A

S. L. Booth, T. Johns, and H. V. Kuhnlein


Difficulties with the published values

Units of expression

Nutrient values of preformed vitamin A and provitamin A can be combined into a single numerical value of vitamin A activity [38]. Originally, the internationally accepted values were IUs [3]. One IU was defined as 0.30 g of all-trans retinol, or 0.60 g of all-trans , -carotene. These units are still found in many food composition tables.

In theory, 1 mmole of all-trans, -carotene should cleave to form 2 mmoles of all-trans retinol [39]. The widely accepted central cleavage theory describes the conversion by the cleavage of the central double bond, whereas the more ambiguous excentric cleavage theory proposes that the vitamin A aldehyde, retinal, is formed by cleavage of one or more of the other double bonds in ,-carotene. The discrepancies in conversion have also been attributed to factors influencing bioavailability and absorption. which in carotenoids are multiple, including the amount of carotenoid in the diet, interactions with other carotenoids, dietary fat and fibre, nutritional deficiencies of zinc and/or protein, and other disease states [8]. The absorption rate for carotene is 20%-50% of that of retinol, which is estimated at 70%-90%, but absorption of the former is less efficient with increasing levels of intake [4]. Other factors that limit the bioavailability of carotenoids relate to the substrate requirements for absorption. Carotenoids can only be absorbed from a micelle in the presence of bile salts, while retinol can also be absorbed from a micelle in the presence of a non-ionic detergent [40]. Retinol is absorbed by diffusion when present in high doses, but is carrier-mediated at low doses. In contrast, carotenoids are absorbed by passive absorption regardless of the concentration.

Given the strong evidence for lower bioavailability, the values for the biological activity of all-trans -carotene were revised. On the basis of rat studies, it was estimated that the other provitamin A carotenoids have 50% of the growth-promoting activity of -carotene [40]. This gave rise to units of expression for vitamin A activity called retinol equivalents, or RE. These are now the internationally accepted units for vitamin A activity, and can be summarized as follows:

1 RE = 1 g all-trans retinol
= 6 g all-trans ,6-carotene
= 12 g other biologically active carotenoids
= 3.33 IU retinol
= 10.0 IU carotene [6]

However, there is still confusion between IU and RE, given the differences in equivalency when converting -carotene to retinol. Use of RE reduces the contribution of provitamin A to total vitamin A activity compared to the system of IU. The formula for conversion is as follows:

RE = IU retionol/3.33+ IU -carotene/10.0 = g retinol
= g -carotene/6 +g other carotenoids/12
(A more thorough presentation of formulae for interconverting units of vitamin A is found in Olson [3].)

Other vitamin A-related compounds exist in dietary sources of preformed A, particularly in fish liver and oils. All-trans dehydroretinol, referred to as vitamin A2 in the older literature [12], is a vitamin A related compound found in freshwater fish flesh and liver and, to a lesser extent, in some marine fish [40, 41]. This compound is estimated to have 40%-50% of the vitamin A activity of all-trans retinol [42]. Likewise. cis isomers of retinol, which can account for up to 35% of preformed vitamin A measured in fish liver oils, have up to 75% relative activity of all-trans retinol. These discrepancies in vitamin A activity are frequently overlooked in food composition literature, although recent studies are adjusting retinol activity values according to differential biological activity [31, 35]. Isomerization, which will be discussed in greater detail later, is also an important issue in quantifying provitamin A activity in processed forms of plant products but has not yet been given adequate attention in the calculation of units for expressing vitamin A activity.


Analytical techniques

Provitamin A

With recent interest in the possible link between cancer and the intake of carotenoids [43-47], an extensive literature has emerged describing the available methods for analysing carotenoids, particularly those by HPLC. Several thorough reviews exist, with descriptions of the theoretical and practical applications of each method [48-50]. The methods can be summarized as follows:

biological methods;

physico-chemical methods

Bioassays are expensive both in time and money, and lack the precision of the physico-chemical methods that have now replaced them. Of the physico-chemical methods, the two of most significance for this discussion, given their extensive use, are column chromatography and HPLC.

Carotenoid analysis is accomplished by extraction, followed by partial purification, separation according to hydroxyl groups, isolation by chromatography, and then measurement by spectral absorption [51]. The AOAC method for carotene analysis is an open-column chromatography method using a magnesium oxide column, which separates carotenoids from xanthophylls on the basis of polarity, followed by visible absorption spectrophotometry [52]. The first fraction eluted is assumed to be -carotene.

Recent studies in carotenoid analyses have revealed that assumptions inherent in the AOAC method are incorrect, so that much of the published provitamin A nutrient data overestimates the true value of certain foods [6]. The assumption that all carotenoid activity measured is that of -carotene can create error, given the lower biological activity of other carotenoids that are eluted in the same fraction [49]. For raw green leafy vegetables, in which the active carotenoids are almost exclusively, (-carotene, the assumptions of the AOAC method are still valid [49, 50]. For food items with mixed carotenoid activity, such as squashes and carrots, -carotene and other carotenes will be measured as -carotene, though they have less vitamin A activity. For this category of foods, inclusion of a step wise gradient in the analytical technique is recommended [6]. In fruits, when xanthophylls are esterified their polarity is reduced so they can elute in the same fraction as -carotene, creating an overestimate of vitamin A activity [51]. The use of a saponification step is recommended for these food items, as it hydrolyses the ester linkage [6, 53]. The saponifiable material is discarded. Plant oils, such as palm oil, also require saponification prior to extraction [49].

Reversed-phase HPLC is rapidly becoming the preferred method for carotenoid analysis, given its flexibility in the identification and quantification of the numerous carotenoids within a single food item. Much of the current literature is a description of evolving methods for the analysis of provitamin A activity in fruits and vegetables [53-59]. While it is argued that HPLC is a superior method [6], the complexity of carotenoids, their isomers, and other chemical substances in foods have so far prevented the development of a single method [51]. Likewise, extraction methods vary with the food substance analysed [57], as does the selection of columns and solvent systems [54], which can create discrepancies in provitamin A values that are not explained by natural sources of variation. Other limiting factors in standardizing analysis of provitamin A activity using HPLC have been reviewed elsewhere [49].

Another major limiting factor for all analytical methods, especially HPLC, is the cost of equipment and solvents, which is prohibitive in most developing regions [49]. A modification of the open column chromatography/visible absorption spectrophotometry method has demonstrated high repeatability and flexibility to adapt to the nature of the food item [60]; this may be a viable alternative for workers in regions that cannot gain access to HPLC.

Preformed vitamin A

Several recent reviews have evaluated the methods available for the analysis of preformed vitamin A in food items [38, 42, 61]. The major categories of methodology can be summarized as follows:

biological methods;
physico-chemical methods

Of these, the Carr-Price reaction is the official AOAC procedure for retinol analysis [52]. Retinol (and its esters) produces a blue colour in reaction with antimony trichloride, so the intensity of the coloured product is used to estimate the retinol concentration of the sample. The reproducibility is acceptable, and, as a consequence, this has been the most common procedure used over the past 40 years [38]. Limitations cited for this method include fading blue colour, use of corrosive reagent, inability to differentiate between retinol derivatives, sensitivity of the reagent to moisture, and interference from carotenoids [42]. If carotenoids are not removed by chromatography, corrections must be made for them.

The use of UV absorbance, also known as the Bessey-Lowry method [6], was an accepted procedure for more than 50 years [38]. The absorbance of the extract is read, the retinol is irradiated with UV light. and then the absorbance of the extract is reread. The difference in absorbance is the concentration of retinol in the sample. This method assumes that only retinol is destroyed at this absorbance and therefore, in the absence of a sensitive spectrophotometer, does not differentiate potential interference from various lipids and fat-soluble vitamins that absorb in the same region of the spectrum. Despite this limitation, which led to the use of correction factors to minimize interference from other compounds, it is argued that use of this method from the 1930s to the 1950s to determine retinol content in oils and concentrates produced some of the most reliable data to date [38]. Validation of the UV method using colorimetric analysis of the same samples of polar bear liver was confirmed in one study, with good agreement between the two methods [37]. As previously discussed, saponification is an essential step for conversion to retinol, with the saponifiable portion being discarded [42]. This procedure also frees the ester form from matrices of stabilized vitamin A products.

More recently, HPLC is appearing in the literature as the preferred method of retinol analysis. Chromatography in itself is not a method to measure retinol; instead it is the procedure for separation, with spectrophotometry and fluorometry used for quantification [61]. The advantages of HPLC are numerous, including versatility, high reproducibility, and greater facility to eliminate interfering substances [38].

However, at present, when methodologies using HPLC are evolving rapidly, standardization is difficult to attain. Coefficients of variation between laboratories can exceed 20%, and such a wide margin of variation creates difficulties in the interpretation of nutrient data for vitamin A activity.


Natural sources of variation

Regardless of the analytical method selected for provitamin A and vitamin A determination, error introduced during the collection and preparation of samples can create large discrepancies in the final nutrient values. Sources of sampling and preparation error have been reviewed [62, 63], and it has been shown that much of the variation is attributable to the nature of the foods being analysed [38].

Heterogeneity in nutrient content is a consequence of numerous factors, including soil pH, amount of rainfall, seasonality, genetic diversity, and the stage of maturation. Moreover, the vitamin is not uniformly distributed within the animal or plant tissue, so the accuracy of the nutrient value is determined in part by the portion size and the number of individual units selected for a representative sample. Bureau and Bushway [64] found that the range in provitamin A values for 22 fruits and vegetables sampled was very large and not consistent across seasons or locations. Nutrient data for green leaves analysed in two different seasons also showed inconsistent variation between seasons [65]. It has been demonstrated that time of marketing affected total vitamin A activity in certain fruits [66]. This variation attributable to different cultivars and handling conditions has been confirmed [67]. In a longitudinal study of nutrient composition fluctuations in carrots, the carotene concentration was found to decrease with delayed sowing [63]. Likewise, the ,-carotene concentration in sweet potato greens cut only one time demonstrated a significant interaction between cultivar and the time of harvest [69]. Sampling considerations also need to include diurnal variation of, -carotene, which has been documented in leafy vegetables [70].

Preformed vitamin A sources also show large ranges of variation. Retinol concentrations in liver oils among fish species can range more than a thousandfold, and mammalian liver retinol concentrations can differ within species more than two hundred-fold [12, 31]. One source of intra-species variation is the age of the animal: young domestic animals have much lower vitamin A liver reserves than their adult counterparts. Age-related differences, however. did not explain the twofold range in preformed vitamin A concentrations in polar bear liver [37]. Dietary differences can also account for differences in retinol concentrations. Yellow fat deposits found in animals may contain large quantities of carotenoids that contribute to the total vitamin A activity [12].


Effects of processing

Carotenoids and retinol are affected by pH, enzymatic activity, light, and oxidation associated with the conjugated double bond system [62]. The chemical changes occurring in carotenoids during processing have been reviewed by Simpson [71]. Fresh plant tissue may contain enzymes that are only activated during, and following, processing [62]. As a consequence of these chemical changes, the preformed vitamin A or provitamin A content of the raw form of a food item may be reduced in food preparation. The most dramatic example of this is found in red palm oil, which in its raw form is considered one of the richest sources of provitamin A [10]. After heating to 200C for 30 minutes, the -carotene content becomes negligible.

Numerous reports document changes in the provitamin A content attributed to various forms of cooking [15, 27, 72-76]. Regardless of the method used, dehydration significantly reduces the carotene content in vegetables, which has implications for storage of seasonally available foods [76, 77]. However, others argue that mangoes that are sun-dried and stored up to six months still have adequate vitamin A activity [25].

Changes in provitamin A content have been documented with traditional methods of preparing leafy vegetables and tubers, but the changes vary with the method used. As a general rule, foods boiled in an open container showed the greatest losses. Whereas one study in Bangladesh reported losses of up to 43% in leafy green vegetables following the traditional method of boiling with subsequent frying [ 15], another study in Indonesia reported negligible losses when sweet potatoes and leafy greens were fried and then boiled [78]. Other authors report either no change or increases in carotene content following certain processing methods [18, 51, 72, 74, 77]. However, values between studies are difficult to compare, given natural sources of variation, different analytical techniques, and an absence of data on the original moisture content.

Sweeney and Marsh [79] reported that processing of fruits and vegetables induced isomerization of -carotene, resulting in an estimated 15%-20% reduction in vitamin A potency in green leafy vegetables, and 30%-35% in yellow vegetables. The vitamin A activity in the raw form of these foods is predominantly all-trans -carotene. With the introduction of HPLC, which offers the sensitivity and reproducibility to isolate and quantify stereoisomers, several recent studies have confirmed that cis isomers exist in human serum and in raw fruits and vegetables [80-84].

Moreover, with increased temperature, the presence of light, and catalysts such as acid, isomerization to the cis form increases [72].

The implication of this isomerization is related to the lower biological activity associated with the cis form, which is estimated to have 38%-53% of the potency of the all-trans form [79]. The reports of an increased carotene content following processing may reflect a loss in soluble solids [74] and an associated increase in concentration of cis isomers, which, in terms of vitamin A potency, would be significantly lower than in the raw form. However, a recent study of Indonesian greens demonstrated that the reduction in vitamin A activity due to cis isomer formation during traditional processing did not exceed 10%, whereas the large variation in cis isomer content in raw vegetables selected from different markets accounted for a reduction of vitamin A activity of up to 12% [78]. This suggests that variation attributable to sampling procedures exceeds that associated with processing foods by traditional methods.

The documentation of processing effects on preformed vitamin A sources is less abundant. Losses of up to 40% in fish sources rich in vitamin A have been reported following boiling [85]. In a study examining traditional foods eaten by people of the Dene nations of the Northwest Territories of Canada, processing of animal and fish livers by baking did demonstrate some decrease in total vitamin A activity, but this was not consistent [35]. Pickling of cheeses with a cow and buffalo milk base led to slight losses in total vitamin A activity, but these were attributable to changes in moisture content during the pickling process [34].


Bioavailability and metabolism

Both nutritive and non-nutritive substances in foods affect absorption, and hence availability and metabolism of both preformed vitamin A and provitamin A [86]. Utilization of retinol and, -carotene is improved with both protein and fat intake [8]. Dietary intake of vitamin E and zinc also enhances utilization of preformed vitamin A and provitamin A. Different isomers and related compounds have demonstrable differences in biological activity. Rat studies indicate that excessive intake of preformed vitamin A decreases -carotene liver stores but increases liver stores of retinol [87]. Excessive intake of either decreases vitamin E status.

The addition of green leafy vegetables to a basal diet of maize and beans increases weight gain in vitamin A-deficient rats, suggesting that provitamin A can have a role in improving vitamin A status [88]. Several metabolic studies involving children document a significant increase in serum vitamin A levels with increased intake of green leafy vegetables [89-92]. This would confirm results from epidemiological studies reporting a negative association between fruit and vegetable intake and xerophthalmia [93-96]. Absorption of, -carotene in the metabolic studies was estimated to range from 61% to 70%. A study by Hussein and El-Tohamy [91] showed that an oral dose of 200,000 IU (21 servings of carrots over a period of 40 days) and an equivalent number of servings of green leafy vegetables were equally effective in raising serum levels after 40 days. Carotene digestibilities (percentage apparent absorption) of 47% and 81% for carrots and spinach respectively were reported following a two-week supplementation study in young males [97]. Lala and Reddy [92] reported increases in serum retinol levels in malnourished children 15 days after treatment with green leafy vegetables. The most dramatic increases were in those children with the lower serum levels at the onset of the study, which confirms the adaptive efficiency of -carotene absorption. Rains-Mariath et al. [11] demonstrated partial or complete regression of clinical signs of xerophthalmia in 10 of 12 children fed Mauritia vinifera Mart. fruit over a period of 20 days.

Two groups [89, 92] noted that respiratory infection and other infectious diseases reduced -carotene utilization, with lower serum levels being measured during the period of illness. This was confirmed by Hamdy et al. [98], who also reported that provitamin A supplementation did not change serum levels in subjects with parasitic infection. While the strong interrelationships between vitamin A deficiency and various infectious diseases are well documented, the actual mechanisms have yet to be clarified [99].

More recent studies focused on the bioavailability of carotenoids suggest that, within a normal population, large inter-individual variations exist in plasma concentrations of ,-carotene [100-102]. In a study of 30 men, a threefold to fourfold inter-individual variation in the efficiency of carotenoid absorption was noted [100]. Eleven days of supplementation with carrots or pure ,-carotene resulted in an increase in plasma levels of -carotene, whereas supplementation with broccoli or tomato juice did not. These authors concluded that plasma carotenoid levels reflected the long-term dietary intake of provitamin A sources. These levels do not appear to fluctuate in plasma levels with the increased intake of a single food item, as demonstrated by low intra-individual variation. Moreover, certain individuals were consistently less efficient in carotenoid absorption than other subjects in the study, which would suggest that plasma levels do not always reflect dietary intake. Significantly higher levels of -carotene absorption in subjects consuming a high-fat diet than in those on a low-fat diet have been reported [101].


Food composition tables

Accurate food composition data are needed for calculation of the vitamin A intake of a population from dietary surveys and for the selection of foods rich in vitamin A for education programmes. Food composition tables contain crude nutrient values from the chemical analyses of foods, with no allowance for the biological utilization of the item [86], so these values are estimates at best. The limitations of vitamin A and provitamin A nutrient values in existing tables have been reviewed [6] but several issues deserve mention.

The different units for expressing vitamin A activity have created confusion. In the recent USDA Handbook No. 8, vitamin A activity is presented as both RE and IU [103]. FAO tables list values in micrograms of retinol or -carotene equivalents [14, 21, 104], as do other food composition tables [16, 105], which allows the user to differentiate between preformed vitamin A and provitamin A activity or to calculate total vitamin A activity. Some tables present total vitamin A activity in RE [106]. Other tables are not as flexible or accurate in terms of vitamin A activity [6]. It has been common practice to assume that 1 IU of retinol is equivalent to 1 IU of -carotene, and then to group the values for vitamin A and provitamin A together [40]. The vitamin A activity values in the Latin American food composition table [107] were recalculated into micrograms of retinol, -carotene, and other carotenoids [19] based on the FAO table that estimated the distribution of provitamin A and preformed vitamin A in different foods [104]. Similarly. other food composition tables have reconverted original nutrient data from IU to RE using this distribution table [14, 21]. Recalculations were done using data from outdated methods in which many of the original data were analysed in the 1940s and 1950s [6].

The AOAC method for provitamin A determination is not adapted to the complexity of carotenoid profiles in certain foods, so many of the values, particularly those in fruits, are overestimates of the actual vitamin A activity. A comparison of values from USDA Handbook No. 8 [103] with HPLC analysis of the same foods revealed substantial differences in fourteen items, with nine overestimated in the handbook and five underestimated [64]. Discrepancies attributable to outdated methods were also demonstrated in the Thai food composition table [108]. Using 108 dietary records of 24-hour recall, total dietary intakes of macronutrients and micronutrients were calculated from nutrient data from the table and compared with direct chemical analyses of the foods using HPLC. Estimated vitamin A intake in RE as calculated from the table was 243% greater than that calculated from the chemical analyses. The greatest discrepancy occurred in vitamin A values for fruit intake—a 203% estimate. Preformed vitamin A sources (eggs and poultry) were in greater agreement, with the overall modification shifting the percentage contribution of total RE to the retinol sources.

As current food composition tables give little information on the carotenoid profile or the stereoisomers of retinol or carotenoids, assumptions need to be made about the contribution of the food item to the total vitamin A activity. These are not made uniformly across all food tables [56]. For example, original data presented as total carotenes have often been converted to IU by the arbitrary division of one-third part -carotene and two-thirds parts other carotenoids [23].

Use of appropriate sampling procedures reduces sampling variance, with the nutrient value representing a mean [38]. Caution must be exercised when comparing values from samples with different sampling and handling procedures [63], yet these differences are inherent in food composition tables. As reviewed by Moore [12], reproductive cycle, age, and the species of fish are sources of marked variation in preformed vitamin A activity of fish oils, but these factors are rarely given consideration in the food tables. When nutrient values from locally analysed foods are not available, food composition tables may use values from other regions, which increases the probability of error, given the geographical variation in nutrient content of foods. This has been demonstrated in provitamin A activity values by comparing Turkish with American dried apricots [109]. The Latin American food composition table [107], for example, includes nutrient values from the USDA Handbook No. 8 [110] and FAO publications of 1949 [111]. However, as mentioned in numerous nutrition studies, a complete database for provitamin A and preformed vitamin A values for locally available food items is often lacking [31, 112].



  1. Vitamin A activity in foods is currently expressed as retinol equivalents (RE). Conversion rates from the carotenoids to retinol adjust for certain factors affecting absorption and utilization, but have yet to address differential vitamin A activity due to isomerization.
  2. HPLC is currently the preferred analytical technique for determination of preformed vitamin A and provitamin A, given the discrepancies in provitamin A values found using earlier techniques.
  3. Natural sources of variation, notably soil pH, rainfall, seasonality, genetic diversity, and stage of maturation, create large ranges in the preformed vitamin A and provitamin A contents of foods consumed in the diet.
  4. Preformed vitamin A and provitamin A contents of foods are reduced during food preparation, although nutrient data are inconsistent among studies. Isomerization of the carotenoids during processing results in lower biological activity, which has important implications for the promotion of foods rich in provitamin A.
  5. Both absorption and utilization of preformed vitamin A and provitamin A are improved with concurrent dietary intake of fat, protein, vitamin E, and zinc. Other physiological and dietary factors may also influence the efficacy of carotenoid and retinol utilization, but the mechanisms have yet to be clarified. This points to the need for complete food intake data, and possibly for the promotion of additional food items rich in other nutrients, notably fat, to optimize the utilization of preformed and provitamin A in natural food sources.
  6. In their current state, food composition tables contain inconsistencies in preformed vitamin A and provitamin A nutrient values. Differential use of units and conversion rates and reliance on data based on outdated analytical techniques limit their use in the identification of vitamin A-rich foods and the calculation of dietary intake of vitamin A, particularly from carotenoid sources.

Contents - Previous - Next