Article Text
PDF
Abstract
Objectives This systematic review with meta-analyses of randomised trials evaluated the preventive effects of vitamin A supplements versus placebo or no intervention on clinically important outcomes, in people of any age.
Methods We searched different electronic databases and other resources for randomised clinical trials that had compared vitamin A supplements versus placebo or no intervention (last search 16 April 2024). We used Cochrane methodology. We used the random-effects model to calculate risk ratios (RRs), with 95% CIs. We analysed individually and cluster randomised trials separately. Our primary outcomes were mortality, adverse events and quality of life. We assessed risks of bias in the trials and used Grading of Recommendations, Assessment, Development, and Evaluations (GRADE) to assess the certainty of the evidence.
Results We included 120 randomised trials (1 671 672 participants); 105 trials allocated individuals and 15 allocated clusters. 92 trials included children (78 individually; 14 cluster randomised) and 28 adults (27 individually; 1 cluster randomised). 14/105 individually randomised trials (13%) and none of the cluster randomised trials were at overall low risk of bias. Vitamin A did not reduce mortality in individually randomised trials (RR 0.99, 95% CI 0.93 to 1.05; I²=32%; p=0.19; 105 trials; moderate certainty), and this effect was not affected by the risk of bias. In individually randomised trials, vitamin A had no effect on mortality in children (RR 0.96, 95% CI 0.88 to 1.04; I²=24%; p=0.28; 78 trials, 178 094 participants) nor in adults (RR 1.04, 95% CI 0.97 to 1.13; I²=24%; p=0.27; 27 trials, 61 880 participants). Vitamin A reduced mortality in the cluster randomised trials (0.84, 95% CI 0.76 to 0.93; I²=66%; p=0.0008; 15 trials, 14 in children and 1 in adults; 364 343 participants; very low certainty). No trial reported serious adverse events or quality of life. Vitamin A slightly increased bulging fontanelle of neonates and infants. We are uncertain whether vitamin A influences blindness under the conditions examined.
Conclusions Based on moderate certainty of evidence, vitamin A had no effect on mortality in the individually randomised trials. Very low certainty evidence obtained from cluster randomised trials suggested a beneficial effect of vitamin A on mortality. If preventive vitamin A programmes are to be continued, supporting evidence should come from randomised trials allocating individuals and assessing patient-meaningful outcomes.
PROSPERO registration number CRD42018104347.
- NUTRITION & DIETETICS
- PUBLIC HEALTH
- Systematic Review
- Primary Prevention
Data availability statement
No data are available. No additional data is available.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
STRENGTHS AND LIMITATIONS OF THIS STUDY
Our work represents a comprehensive review of the topic, including 120 randomised clinical trials with more than 1.6 million participants, which increases the precision and power of our analyses.
We conducted a thorough review following Cochrane, implementing findings of methodological studies.
In our main analysis, we pooled data from participants of different age strata, coming from different countries, with different socioeconomic status, and participants can differ in their susceptibility to vitamin A deficiency.
Introduction
Vitamin A is essential for vision, regulation of growth, reproduction, immune response and in maintaining tissue integrity.1–4 Our body cannot synthesise vitamin A, and, therefore, we must obtain it through our diet. A balanced diet provides sufficient amounts of the recommended dietary allowances for vitamin A.5 Low vitamin A intake during infancy and childhood, especially in low-income and middle-income regions, raises the risk of vitamin A deficiency.6 7 Vitamin A deficiency may lead to infections, anaemia, growth retardation and xerophthalmia.3 Vitamin A deficiency in adults is rare.5 However, adults in high-income countries often use dietary supplements of vitamin A and increase the risk of intakes above tolerable upper intake levels.5
Excessive vitamin A intake from food or supplements may lead to vitamin A toxicity.8 9 Supplementing children in India with massive doses of vitamin A caused illness and death.10 11 Neonatal vitamin A supplementation seemed to be associated with a cluster of deaths and poor early growth among low birthweight boys.12 High doses of vitamin A seem potentially harmful in adults.13 14 WHO recommends vitamin A only for children aged 6 months to 59 months,15 supported by a Cochrane review focusing only on this age stratum.16 How the division into the different age groups was developed for the Cochrane reviews and the WHO recommendations is not clear.17–19 Why should the effect of vitamin A be different in children aged 6–59 months compared with other age groups?
Our aim with the present systematic review was to evaluate the preventive effects of vitamin A supplements versus placebo or no intervention on all-cause mortality and other clinically relevant outcomes, in people of any age.
Methods
We used Cochrane methods20 21 described in our published protocol.22
Inclusion criteria
We included randomised preventive trials in children (<18 years) and adults (≥18 years) who were either healthy or recruited among the general population (primary prevention); or diagnosed with a specific disease in a stable phase (secondary prevention, eg, low birthweight neonates, vitamin A deficient children, anaemic children, elderly institutionalised people or adult male alcoholics).
Exclusion criteria
We excluded tertiary prevention trials in which vitamin A was used to treat a specific disease (eg, vitamin A deficiency, measles, diarrhoea, bronchitis, bronchiolitis, meningitis, schistosomiasis, tuberculosis, pneumonia, malaria or HIV), malignant diseases and hereditary diseases (eg, Down’s syndrome, cystic fibrosis or biliary atresia). We excluded trials in pregnant and postpartum (lactating or non-lactating) women because of their specific physiological states.
Types of interventions
Vitamin A supplementation (the experimental intervention) could have been administered orally or parenterally at any dose, duration and regimen. Placebo or no intervention was the control interventions.
Types of outcome measures
Our primary outcomes were all-cause mortality; serious adverse events and health-related quality of life.22 Our secondary outcomes were adverse event(s) not considered serious; and blindness.22
Search methods for identification of trials
We searched for trials in the Cochrane Central Register of Controlled Trials (CENTRAL) in the Cochrane Library; MEDLINE; Embase; LILACS; Science Citation Index Expanded and Conference Proceedings Citation Index Science23 (online supplemental table 1); as well as homepages and databases of regulatory authorities. The last searches were conducted from 1946 to 16 April 2024. We scanned reference lists. We did not restrict our searches to year, language or type of publication.
Supplemental material
Data collection and analysis
One review author (GB) performed searches in the electronic databases to identify the trials of possible interest. Three review authors (GB, DN and MB) independently participated in the manual searches and identified trials eligible for inclusion from the search results. To determine if a study should be assessed further, four review authors (GB, CSP, NS and SKK) independently scanned the abstract, title or both sections of every record retrieved. In case of disagreements which we could not solve among ourselves, we consulted another author (CG). We assessed the full text of all potentially relevant publications. One review author (GB) listed the excluded studies with the reason for exclusion.
Assessment of risk of bias in included trials
We assessed the risk of bias of each included trial with risk of bias tool 1 according to the recommendations in the Cochrane Handbook for Systematic Reviews of Interventions.20 To assess the risk of bias in trials that randomised participants individually, we used the definitions shown in online supplemental table 2.21 To assess the risk of bias in cluster randomised trials, we did similarly plus assessed additional bias risk domains such as recruitment bias; baseline imbalance; loss of clusters; incorrect analysis and comparability with individually randomised trials.24 We also assessed the influence of vested interests on our results.25 We performed adjusted-rank correlation26 and regression-asymmetry27 tests for detection of publication bias.
Assessment of heterogeneity and inconsistency
We assessed heterogeneity by visual inspection of the forest plots, and by using a standard χ2 test and a significance level of α=0.1, in view of the low power of such tests. We calculated inconsistency (I2, which describes the percentage of variation across studies that is due to heterogeneity rather than chance.28
Data synthesis
Meta-analysis
We conducted our meta-analyses following Cochrane methods.21 29 We used Review Manager (RevMan)30 to analyse data, using the random-effects model in our primary meta-analyses.21 31 32
Review Manager V.5 does not include trials with zero events in both intervention groups when calculating relative risk ratios (RRs).30 We assessed the influence of trials with zero events in the treatment, control or both groups by recalculating the random-effects model meta-analyses with 0.01 as the empirical continuity correction33 34 using Trial Sequential Analysis V.0.9.5.10 beta.35 36 We also used STATA V.8.2 (STATA) and Sigma Stat V.3.0 (SPSS) for the statistical analyses.
We analysed the cluster randomised trials following Cochrane.24 A major implication of cluster randomisation is that the outcome data on individuals within the same cluster tend to be correlated. The statistical measure of the degree of correlation is known as ‘intracluster correlation coefficient’ (ICC).37 Where results did not control for clustering, we contacted trial authors to request an estimate of the ICC. If trial authors were unable to provide the ICC, we calculated it by ourselves, using design effects. We used previously reported design effects38 to calculate ICCs39–45 as well as ICC estimates obtained from a previously published Cochrane review.19 We used the ICC to obtain an effective sample size for each cluster randomised trial.
We compared the intervention effects in subgroups of trials using the method described by Borenstein et al 46 and implemented in Review Manager V.5 for all types of meta-analyses.
Trial sequential analysis
We used trial sequential analysis because cumulative meta-analyses are at risk of producing random errors due to sparse data and repetitive testing of the accumulating data.35 36 47–54 A more detailed description of trial sequential analysis can be found at www.ctu.dk/tsa/.35 36
GRADE assessments
We used GRADEpro to construct a ‘Summary of findings’ table.55 We assessed the certainty of evidence of all our review outcome results. The certainty of evidence is categorised into ‘high’, ‘moderate’, ‘low’ or ‘very low’.56
Patient and public involvement
Patients and/or the public were not involved in the design, conduct, reporting or in dissemination plans of this research.
Results
Results of the search
Searching databases and reference lists of articles of interest, we retrieved 10 523 references. We excluded 2166 duplicates and 8052 clearly irrelevant references by reading the abstracts. Accordingly, 305 references were retrieved for further assessment. Of these, we excluded 175 references because they did not fulfil our review protocol inclusion criteria. In total, 130 randomised trials described in 281 publications fulfilled our inclusion criteria (figure 1).12 39–45 57–178 The trials randomised a total of 1 674 257 participants. Four trials with 1977 participants reported about 16 deaths, without specifying in which intervention group the deaths occurred.73 134 163 167 Two trials with 239 participants in low birthweight neonates77 88 and four trials in adults with 730 participants169–172 did not report mortality data. The authors of these 10 trials did not reply to our request for additional information. Accordingly, 120 trials with 1 671 672 participants provided data for our analyses of mortality. All trials were published in English.
PRISMA flow diagram (Date of search 1946 to 16 April 2024). PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses.
Trial characteristics
Randomised individuals were the unit of allocation in 105 trials including 239 974 participants and clusters in 15 trials including 1 431 698 participants.40–45 58 64 95 101 102 124 150–152
Individually randomised trials
The characteristics of the included individually randomised trials are shown in online supplemental table3–5. The following is a succinct overview.
88 trials used a parallel-group design, 15 trials used the 2×2, 1 trial a half-replicate of 2×2×2×2,75 and one trial the 2×2×2×2 factorial design (online supplemental table 3).145
The trials were conducted in Asia (n=40), Africa (n=18), Europe (n=16), North America (n=16), South America (n=10) and Oceania (n=4). One trial was conducted in centres in Ghana, India and Peru.157 Most of the trials came from lower-income and middle-income countries (online supplemental table 4).
All 105 trials were published from 1976 to October 2021.
In 41 trials (39%), vitamin A was provided free of charge by pharmaceutical companies. In the remaining 64 trials, funding was not reported (online supplemental table 4).
Participants
There were 239 974 participants randomly assigned in the 105 included trials. The number of participants in each trial ranged from 30 to 44 984 participants (median 232). The age range of participants was from 1 day to 103 years. The mean proportion of females was 49% in the 86 trials reporting sex (online supplemental table 3).
16 trials included low birthweight neonates12 71 72 83 100 112 125 126 136 147 153 156 161 165 166 177; 13 trials included neonates39 70 84 87 97 116–118 123 151 155 157 176; 5 trials included infants 1–6 months of age65 67 103 113 141; 29 trials included children 6–59 months of age60 66 68 69 74 78–80 85 86 90 91 98 99 106–108 110 111 132 135 139 142–144 146 148 149 154; 6 trials included children up to 10 years of age81 82 119 130 131 162; 9 trials children 5–18 years of age3 5 10 12 59 62 114 128 140 159 160 and 27 trials included adults.57 61 63 75 76 89 92–94 96 105 109 115 120 121 127 129 137 138 145 158 164 168 173–175 178
Experimental interventions
Vitamin A was administered orally in 98 trials, intramuscularly in 6 trials,72 100 112 125 126 153 and orally and intramuscularly in one trial.166 Vitamin A was used in different dosing regimens (online supplemental table 5).
Comparator interventions
99 trials used placebo and 6 trials used no intervention in the control group (online supplemental table 3).
Concomitant interventions
42 trials did not use concomitant interventions. In 63 trials, participants were additionally treated with different vitamins, minerals, antiparasitic drugs or vaccines (online supplemental table 5).
Main outcomes
The main outcomes in the trials were mortality and morbidity (online supplemental table 4).
Risk of bias in included trials
14/105 trials (13%) with a sample size of 132 248 participants (55.1%) were considered at low risk of bias, that is, they had adequate generation of the allocation sequence, allocation concealment, blinding of caregivers and participants, blinding of outcome assessors and they were free of selective reporting and other biases.39 63 70 71 87 112 116 117 123 156 165 177–179 The remaining 91 trials with effective sample size of 107 726 participants had 1 or more unclear or inadequate bias risk domains and were considered at high risk of bias (figure 2).
Risk of bias summary: review authors’ judgements about each risk of bias item for each included trial.
Visual inspection of the funnel plot did not suggest potential publication bias (asymmetry) (figure 3). The adjusted-rank correlation test (p=0.85) found no significant evidence of bias while the regression asymmetry test (p=0.031) found significant evidence of bias on the intervention effect of vitamin A supplementation.
Funnel plot of comparison of vitamin A supplementation versus placebo or no intervention on all-cause mortality in 120 trials at low or high risk of bias in children and adults.
Effects of vitamin a intervention
All-cause mortality
Vitamin A supplementation had no effect on mortality (RR 0.99, 95% CI 0.93 to 1.05; I²=32%; p=0.19; 105 trials; 239 974 participants; moderate certainty of evidence) (figure 4; table 1). Trial sequential analysis on these trials revealed that the cumulative Z-curve crossed the trial sequential monitoring area for futility after the 20th trial (online supplemental figure 1).
Intervention effect of vitamin A supplementation versus placebo or no intervention on all-cause mortality in individually randomised trials.
Summary of findings of systematic review of vitamin A supplementation versus placebo or no vitamin A
Sensitivity analyses
All-cause mortality according to bias risk of trials
Vitamin A supplementation did not affect mortality in trials at low risk of bias (RR 1.01, 95% CI 0.96 to 1.07; I²=0%; p=0.71; 14 trials; 132 248 participants) (figure 4). Trial sequential analysis on these trials revealed that the cumulative Z-curve crossed the trial sequential monitoring area for futility after the fourth trial (online supplemental figure 2).
Vitamin A supplementation did not affect mortality in trials at high risk of bias (RR 0.94, 95% CI 0.85 to 1.03; I²=40%; p=0.19; 91 trials; 107 726 participants). The intervention effect in the latter trials did not differ significantly from that in trials at low risk of bias (χ2=1.77; p=0.18) (figure 4).
All-cause mortality according to vested interests
Vitamin A did not affect mortality in trials without vested interests (RR 0.98, 95% CI 0.91 to 1.06; I²=18%; p=0.63; 65 trials; 165 943 participants), nor in the trials with vested interests (RR 1.00, 95% CI 0.91 to 1.10; I²=39%; p=0.95; 40 trials; 74 031 participants). The two effect estimates did not differ significantly (χ2=0.06; p=0.81) (online supplemental figure 3).
Subgroup analyses
All-cause mortality in different age groups.
Children at any age
Vitamin A had no effect on mortality in children (RR 0.96, 95% CI 0.88 to 1.04; I²=24%; p=0.28; 78 trials; 178 094 participants) (online supplemental figure 4).
Vitamin A did not affect mortality in trials at low risk of bias (RR 1.01, 95% CI 0.95 to 1.07; I²=0%; p=0.75; 12 trials; 131 250 participants). Vitamin A reduced mortality in trials at high risk of bias (RR 0.79, 95% CI 0.66 to 0.94; I²=18%; p=0.007; 66 trials; 46 844 participants). The effect estimates differed significantly (χ2=7.01; p=0.008) (online supplemental figure 4).
Low birthweight neonates
Vitamin A did not affect mortality of low birthweight neonates (RR 0.99, 95% CI 0.84 to 1.17; I²=0%; p=0.95; 16 trials; 4216 participants) (online supplemental figure 5).
Vitamin A did not affect mortality of low birthweight neonates in trials at low risk of bias (RR 1.05, 95% CI 0.83 to 1.32; I²=0%; p=0.68; 5 trials; 2287 participants), nor in trials at high risk of bias (RR 0.94, 95% CI 0.75 to 1.19; I²=0%; p=0.61; 11 trials; 1929 participants). The effect estimates did not differ significantly (χ2=0.43; p=0.51) (online supplemental figure 5).
Neonates
Vitamin A did not affect mortality of neonates (RR 0.98, 95% CI 0.89 to 1.08; I²=46%; p=0.72; 13 trials; 146 676 participants) (online supplemental figure 6).
Vitamin A did not affect mortality of neonates in trials at low risk of bias (RR 1.01, 95% CI 0.95 to 1.07; I²=0%; p=0.82; 7 trials; 128 963 participants), nor in the trials at high risk of bias (RR 0.70, 95% CI 0.40 to 1.22; I²=48%; p=0.21; 6 trials; 17 713 participants). The effect estimates did not differ significantly (χ2=1.63; p=0.20) (online supplemental figure 6).
Infants 1–6 months
Vitamin A did not affect mortality in trials including infants of 1–6 months (RR 0.96, 95% CI 0.62 to 1.48; I²=0%; p=0.84; 5 trials; 1487 participants). All trials were at high risk of bias (online supplemental figure 7).
Children 6–59 months
Vitamin A reduced mortality in trials including children 6 to 59 months old (RR 0.74, 95% CI 0.56 to 0.98; I²=0%; p=0.03; 29 trials; 20 834 participants). All trials were at high risk of bias (online supplemental figure 8).
Children 5–18 years
The evidence is very uncertain about the effect of vitamin A on mortality in trials including children 5–18 years old (RR not estimable; 9 zero-event trials; 2602 participants). All trials were at high risk of bias.
Adults
Vitamin A did not affect mortality in trials including adults (RR 1.04, 95% CI 0.97 to 1.13; I²=24%; p=0.27; 27 trials; 61 880 participants (online supplemental figure 9).
Vitamin A had no effect on mortality in trials at low risk of bias (RR 1.99, 95% CI 0.60 to 6.57; p=0.26; 2 trials; 998 participants), nor in the trials at high risk of bias (RR 1.04, 95% CI 0.96 to 1.12; I²=25%; p=0.30; 25 trials; 60 882 participants). The two effect estimates did not differ significantly (χ2=1.13; p=0.29) (online supplemental figure 9).
Further sensitivity and subgroup analyses
Please see online supplemental data to manuscript for sensitivity analyses of vitamin A on all-cause mortality, taking trials with zero events into account and subgroup analyses according to the administeed control intervention in the trials, placebo or no intervention; trials comparing primary to secondary prevention; trials administering high-dose vitamin A compared with trials administering low-dose vitamin A; and trials administering vitamin A singly compared with trials administering vitamin A combined with other vitamins or minerals (online supplemental figures 10–14).
Serious adverse events
None of the trials reported serious adverse events other than mortality.
Health-related quality of life
None of the trials reported data on health-related quality of life (or health economics).
Non-serious adverse events
Several adverse events, defined as non-serious, in children were reported (eg, transient bulging fontanelle, fever, vomiting, diarrhoea, convulsions, inability to suck or feed, excessive crying, jaundice, eye infection, ear infection, skin infection, umbilical infection, respiratory infection, headache, loose motion, unconsciousness, irritability, difficulty in breathing, runny nose, cough and headache), as well as in adults (alopecia, cheilitis, dry skin, dysuria, epistaxis, headache, myalgia, nausea and dryness of the mouth). Vitamin A may increase the instances of bulging fontanelle in neonates and infants (RR 2.03, 95% CI 1.26 to 3.41; I²=82%; p=0.004; 9 trials; 110 774 participants; low certainty evidence).
Cluster randomised trials
Randomised clusters were the unit of allocation in 15 trials including 1 431 698 participants.40–45 58 64 95 101 102 124 150–152 All cluster randomised trials were assessed at high risk of bias. A detailed description of the characteristics of the included cluster randomised trials is presented in online supplemental table 6–8.
Effects of vitamin A intervention
Vitamin A supplementation seemed to reduce mortality in cluster randomised trials (0.84, 95% CI 0.76 to 0.93; I²=66%; p=0.0008; 15 trials; 364 343 participants; very low certainty) (online supplemental figure 15). This estimate differed significantly from the estimate of all-cause mortality in individually randomised trials (χ2=7.29; p=0.007).
We performed a sensitivity analysis for all-cause mortality using ICCs of 0.002 and 0.02 to calculate the effective sample sizes for each cluster randomised trial (online supplemental table 9). The random-effects model RRs for the two ICCs of 0.002 and 0.02 were not noticeably influenced, apart from the wider CIs with the larger ICCs (online supplemental table 9).
Blindness and other outcomes
The effect of vitamin A supplementation on blindness is uncertain (RR 0.88, 95% CI 0.72 to 1.07; p=0.19; 1 trial; 28 753 participants; low certainty evidence).
None of the cluster randomised trials reported on our remaining primary or secondary outcomes.
Discussion
Our systematic review contains several major findings. Mortality in trials using individual randomisation, offering the fairest comparison, especially in trials at low risk of bias, is likely not affected by vitamin A. The evidence about the effect of vitamin A on mortality ensuing from cluster randomised trials is very uncertain. Vitamin A did not affect mortality in children and adults. Subgroup analyses in different age groups suggested that vitamin A reduced mortality in trials including children 6–59 months old, but data from trials at low risk of bias, in the specific age group, are needed to conclude with greater certainty. Vitamin A increased bulging fontanelle in neonates and infants and caused other non-serious adverse events. However, based on the certainty of evidence, we are uncertain in the result. We are uncertain whether vitamin A supplementation influenced blindness under the conditions examined. It has been hypothesised that vitamin A supplementation may have beneficial effects on outcomes like child growth and development, immunity and morbidity, but the results of randomised trials have been equivocal.42 74
Certainty of the evidence
We followed our published protocol.22 Our work represents a comprehensive review of the topic, including 120 randomised clinical trials with more than 1.6 million participants, which increases the precision and power of our analyses. Previous meta-analyses of preventive trials of vitamin A supplements have included substantially less information, largely due to their focus on selected age strata.180 181 We conducted a thorough review following Cochrane,21 implementing findings of methodological studies.182–187 We also conducted trial sequential analyses to control the risks of random errors in the cumulative meta-analyses.
The certainty of the evidence in individually randomised trials likely showing no effects of vitamin A supplements on mortality was moderate because of risks of bias. We found comparable results in the individually randomised trials at low risk of bias. Contrary, in the cluster randomised trials, a beneficial effect of vitamin A supplementation on all-cause mortality was only supported by very low certainty evidence which means that we do not know if this result is true. Cluster randomised trials are known to be less reliable compared with individually randomised trials (please see below).
Limitations
As with all systematic reviews, our findings and interpretations are limited by the quality and quantity of available evidence on the effects of vitamin A supplements on mortality. The examined populations varied. In our main analysis, we pooled data from participants of different age strata, coming from different countries, with different socioeconomic status, that can differ in their susceptibility to vitamin A deficiency.
Most of the included trials were considered at high risk of bias, which undermines the validity of their and our results.182–187 Other types of bias like bias from trials with deficiencies in the trial design, small trial bias, vested interests, etc might have influenced our results too (eg, https://www.babymilkaction.org/archives/31386; https://www.gava.org/).
Agreements and disagreements with other studies or reviews
Vitamin A deficiency (defined as serum retinol <0.7 µmol/L or ≤20 µg/dL) may cause night blindness and increase the risk of morbidity and mortality from infections.15 There are three ways to prevent vitamin A deficiency in affected populations: better access to vitamin A-rich foods; fortification of a staple food with vitamin A (where food variety is poor) or periodic delivery of vitamin A supplements.188 189
The WHO recommends vitamin A supplementation in a single dose of 100 000 international units (IU) for infants 6–11 months of age, and 200 000 IU every 4–6 months for children 12–59 months of age in countries where vitamin A deficiency is considered a public health problem (prevalence of serum retinol <0.7 µmol/L ≥20%).15 WHO does not recommend vitamin A supplementation for other age groups.15 The current WHO recommendations for vitamin A supplementation in children15 190 191 are based mainly on the results of some of the older cluster randomised trials in children with flaws in their design and at high risk of bias.41–43 45 124 Extremely positive results of some of these trials have previously been questioned.98 192 Their results drove the final results of subsequent meta-analyses180 181 and of Cochrane reviews.17–19 In contrast, later individually and cluster randomised trials with proper methodology found neutral effects of vitamin A.
Recent literature debates different types of biases that may influence the results of cluster randomised trials.193–195 The comparability of intervention groups is challenged in cluster randomised trials because groups of participants rather than the participants themselves are randomised. The chronology of cluster randomised trials compromises allocation concealment (ie, clusters are recruited and randomised, and then participants are recruited), which can induce imbalances between groups. We lack statistical methods to handle non-recruited participants in cluster randomised trials. Therefore, the principle of intention to treat is also challenged in cluster randomised trials. Consequently, we analysed individually and cluster randomised trials separately. In our review, vitamin A supplementation likely does not reduce mortality in individually randomised trials, and the evidence is very uncertain about the effect of vitamin A supplementation on mortality in cluster randomised trials.
We found neutral effects of vitamin A supplementation vs control on mortality in low birthweight neonates, termed neonates, infants 1–6 months of age, children 5–18 years of age and adults. Our results concur with the results of other systematic reviews and meta-analyses.17 196–198 Moreover, they support recent opinions that universal distribution of high-dose vitamin A to children should cease.199 200
Generally, we found neutral effects of vitamin A supplementation in children of any sex on mortality. However, vitamin A might have beneficial effects in some subgroups and harmful effects in others. Benn et al found some evidence that vaccination status and sex were potential effect modifiers of vitamin A.201 They suggested that vitamin A supplementation in children is reassessed in sufficiently powered randomised trials to detect effect modification by vaccination status, sex and other potential effect modifiers.201 A Cochrane review, on vitamin A supplementation for preventing morbidity and mortality in children from 6 months to 5 years of age, includes subgroup analyses of vitamin A supplementation in boys compared with girls.19 The authors found neutral effects of vitamin A supplementation on mortality in both subgroups. The test for subgroup differences was not significant (p=0.22).19 A recent article also argues that the time is not right for a change to the vitamin A supplementation programme for children in India202 mainly because the diets of children under 5 years old in India are grossly deficit in vitamin A against their recommended dietary allowances, as their diets are predominantly from plant sources.203 Our review showed that vitamin A supplementation did not reduce mortality in trials administering high-dose vitamin A, nor in trials administering low-dose vitamin A.
Our systematic review included 16 randomised trials with more than 4000 participants in low birthweight neonates. We found a neutral effect of vitamin A supplementation on mortality. Other systematic reviews that even speculate that vitamin A might have beneficial effects on survival in this age group included a significantly smaller number of trials.204 205 A recent umbrella review of systematic reviews and meta-analyses, dealing with interventions to prevent bronchopulmonary dysplasia in preterm neonates, concluded that vitamin A supplementation is not a recommended prevention strategy because of the possibility of increased risk of mortality.206 There are speculations in the literature that neonatal vitamin A supplementation reduces mortality in Asia, but not in Africa where it has almost detrimental effects.207 208 Before accepting such regional effects, more likely reasons, such as risks of bias and design errors, should be considered.
Approximately half a billion vitamin A capsules are manufactured yearly and distributed across 100 countries worldwide. However, we lack evidence that high doses of vitamin A result in sustained shift in serum retinol levels and in an effect on the prevalence of vitamin A deficiency in children.209 International funded programmes are not always driven by science. They may be manipulated by policy-makers and influenced by vested interests.210 211 Furthermore, strong WHO recommendations are frequently based on low-quality evidence, particularly regarding child health.212 213 The sustainability of the WHO recommendation for vitamin A supplementation in 6–59 months old children has been repeatedly questioned. A recent study found that even in limited resource settings, the combination of local foods provides 100% of the recommended daily allowance for vitamin A among children 6–23 months old in Ethiopia.214 Vitamin A liver stores were positively associated with breast feeding but not with vitamin A supplementation in Senegalese urban children 9–23 months old.215 Three recent studies raise concerns of vitamin A excess in children when dietary intake, food fortification and vitamin A supplementation are considered together.216–218 Acute and chronic hypervitaminosis A were associated with suboptimal anthropometric measurements in South African children.219 Furthermore, data on the population vitamin A status are limited or absent. Most of the 82 countries that implemented vitamin A supplementation programmes had no vitamin A deficiency data, or the data they had were more than 10 years old.189 Recently, a meta-analysis of five trials of vitamin A supplementation in Indian children showed no significant effect on mortality.220 Another systematic review found that excessive vitamin A supplementation increased the incidence of acute respiratory tract infections.221
Trials assessing vitamin A supplementation in adults have also produced controversial evidence, associated with potential detrimental effects.13 14 222 A recent systematic review did not find beneficial effects of oral vitamin A supplementation for prevention of viral infections.223 However, the review found encouraging results for the management of human papilloma virus lesions and some measles-related complications.223 Today, more than one-half of adults in high-income countries ingest dietary supplements,5 most frequently in the form of multivitamins with or without minerals.224 When combined with dietary intake, the total intake of vitamin A supplement users in the USA exceeds 100% of the estimated average requirement.224 Consequently, vitamin A may provide benefit, but it may also precipitate harm in adults. Some of the analyses of our systematic reviews showed an association between vitamin A supplements, given as primary or secondary prevention, and increased risk of mortality in adults.222 225
The adequacy of current dietary reference intakes (DRIs) for vitamin A was recently questioned.226 DRIs for vitamin A were developed in 2001 using very sparse data for children and adults.227 DRI values for adults were developed using data from four studies with only 13 participants.227 DRIs for children were developed using limited data for infants and extrapolation of data from adults.227
The assumed cut-off for blood retinol (<0.7 µmol/L) that is used to define vitamin A deficiency is based on a small number of studies.228 There are also ethnic and sex differences in serum retinol levels.229 Afro-Americans and Asian-Americans have lower retinol concentrations in blood compared with Caucasians.229 Adult females have blood retinol concentrations lower than males.230 A recent systematic review analysed which cut-off value should be used for vitamin A deficiency in children aged 3–10 years.231 The results showed that blood concentrations of vitamin A had low accuracy to discriminate the outcomes related to vitamin A deficiency (ie, xerophthalmia, immune dysfunction, impaired growth, anaemia) in children of this age group.231 Therefore, we still lack evidence about the optimal vitamin A status.
Conclusions
Based on evidence with moderate certainty, vitamin A supplementation did likely not affect mortality in individually randomised trials. The evidence is very uncertain about the effect of vitamin A supplementation on mortality in cluster randomised trials. Vitamin A increased bulging fontanelle of neonates and infants.
Data availability statement
No data are available. No additional data is available.
Ethics statements
Patient consent for publication
Acknowledgments
We thank Janus Christian Jakobsen, Chief Physician of the Copenhagen Trial Unit, for his expert comments during preparation of the protocol and the review. We are grateful to the authors of the trials for the information on the trials they were involved in. To the Copenhagen Trial Unit, Centre for Clinical Intervention Research, The Capital Region, Copenhagen University Hospital−Rigshospitalet, Copenhagen, Denmark.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Contributors GB and CG were responsible for the study concept and design. GB, DN and CG were responsible for the drafting of the manuscript. MB, CSP, NJS and SKK made several critical revisions and provided professional and statistical support. All authors worked on the review manuscript, fulfilling the prespecified tasks in the methods section of the published protocol. All authors read, revised and approved the final manuscript. GB is responsible for the overall content as guarantor.
Funding This work was supported by Copenhagen Trial Unit, Centre for Clinical Intervention Research, The Capital Region, Copenhagen University Hospital–Rigshospitalet, Copenhagen, Denmark and the Ministry of Education, Science and Technological Development of Republic of Serbia (Grant No: 451-03-66/2024-03/200113).
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.