Summary of main results
Vitamin A in a single dose for treating measles in children was not associated with a reduced risk of mortality. Two doses of vitamin A (200,000 international units (IUs) on consecutive days) reduced the mortality in children aged less than two years and pneumonia-specific mortality. There was no significant reduction in the risk of mortality in the vitamin A group when all the studies were pooled (RR 0.70; 95% CI 0.42 to 1.15). Two doses of vitamin A reduced the incidence of croup but not pneumonia morbidity, nor diarrhea morbidity. However, the mean duration of pneumonia, diarrhea and fever in the intervention groups were shorter, and the mean number of days in hospital in the intervention groups was less.
The quality of most of the trials included in this review is high. The factors included in the subgroup analyses of dose, formulation, setting and age were highly correlated and three studies (Barclay 1987; Coutsoudis 1991; Hussey 1990) were strongly represented in these analyses.
Dose and formulation
This review demonstrates that vitamin A administered to children with measles and receiving standard treatment was associated with a reduction in mortality when children were under the age of two, hospitalized and the dose (200,000 IU) was repeated on the second day. The evidence partly supports the WHO recommendation of two 200,000 IU doses.
Although the data do not allow us to examine the individual effects of dose and formulation, these are issues that need to be considered. Vitamin A preparations in oil and in water are different in terms of their action in the body over a period of time because of differences in the processes of absorption, distribution, localization in tissues, bio-transformation and excretion. Water-based vitamin A preparations lead to greater absorption, which results in higher serum retinol levels. The oil-based preparation is more stable, readily available and costs less. For these reasons it is the latter that is recommended by WHO.
The Coutsoudis study and others (Inua 1983; Markowitz 1989; Reddy 1986) support the finding that serum retinol concentrations are lowered during measles. In Coutsoudis' study (Coutsoudis 1991) the supplemented group had significantly higher concentrations than the placebo group, which indicates that the liver stores were not depleted but that there was temporary impairment of mobilization and increased utilization of vitamin A.
The children in the Rosales (Rosales 1996) and Ogaro (Ogaro 1993) studies may not have benefited from receiving vitamin A oil-based preparations in a single dose (200,000 IU) as this might not have been sufficient to reverse the hyporetinemia occurring during measles; the dose may have been stored, mostly in the liver. Rosales reported a 70% increase in serum retinol after a single dose of oil-based vitamin A. The Rosales study was a community-based study and, therefore, the protective effect of vitamin A may not have been as great as seen in the more severe hospital-based cases.
The results in this review confirm that two doses of vitamin A (200,000 IU) are associated with reductions in the risk of overall mortality and of pneumonia-specific mortality. In 1991, Rosales (Rosales 1996) came to the same conclusion as did Sommer, who suggested that it was prudent to follow the double-dose schedule already proven in the Barclay, Hussey and Coutsoudis trials rather than the single dose recommended by WHO at that time. Doubling the WHO dose was also advocated by Chan (Chan 1990) and Hussey (Hussey 1997). Although use of two doses and the water-based product was associated with a greater reduction in risk of mortality, no recommendation can be made as to whether a single dose of water-based preparation would have a similar benefit as no studies have been conducted looking at the effect of a single dose of water-based vitamin A as compared to two doses. Therefore, single-dose, water-based and oil-based preparations need to be compared to two-dose schedules. The trade-off of using high-dose, oil-based vitamin A versus a water-based formula has to be viewed in terms of the advantages of each product. Although the water-based product may be associated with greater mortality reductions the advantage may be offset by its lower stability, higher cost and non-availability.
One study (Barclay 1987) used two doses of oil-based vitamin A and the effect on overall mortality was not significant on its own, except for children under the age of two years. The evidence for oil-based vitamin A having a protective effect on mortality was demonstrated when an old study by Ellison with a lower quality score was included as part of the sensitivity analysis. Although this study used very small doses of vitamin A (3000 IU for seven days) the supplemented group had statistically significant reductions in risks of mortality, even in the absence of antibiotics and immunization. This study was not randomized and two separate wards were allocated to receive the placebo or vitamin A supplementation. The participants in this study could be comparable to the African children enrolled in the other five studies almost 60 years later as the case-fatality rates in the Ellison study were very similar, and in some cases lower than the case-fatality rate in the placebo and supplemented groups in some more recent studies.
The effect of vitamin A was more pronounced in children under the age of two years as a greater reduction in the risk of mortality was observed in this age group. This was seen across all studies but more so in the studies that used the two-dose regimen (Barclay 1987; Coutsoudis 1991; Hussey 1990). In children under the age of two years formulation did not make any difference as the oil-based product was associated with a statistically significant reduction in the risk of mortality and the water-based vitamin A effect almost reached statistical significance. The study by Markowitz et al (Markowitz 1989) highlighted the fact that children aged less than two years of age with low vitamin A levels had a higher risk of dying than those with higher levels; the number of children in the age group older than two years were too few to detect any statistically significant difference.
Case-fatality rate in country of study
As the studies using two doses (Barclay 1987; Coutsoudis 1991; Hussey 1990) were from areas where case-fatality was more than 10% it is important to be careful in generalizing the results. It raises the issue of whether the decrease in mortality was a result of the higher dose, or whether the vitamin A supplementation in higher case-fatality areas had a greater effect, as there was a greater potential for mortality decline in those populations. It may be possible that there would be a decline in mortality even with a single dose of vitamin A in high case-fatality areas and this needs to be further explored. Although in South Africa the measles case-fatality rate was greater than 10% in hospitals, Coutsoudis had low case-fatality rates in both the vitamin A and control groups. She remarked that this could be attributed to the absence of emergency and malnourished cases.
Hospital versus community studies
The protective effect of vitamin A supplementation was seen only in hospitalized children. Hospitalization may be a measure of severity of illness. There is the possibility that more severe clinical cases of measles are more likely to benefit from vitamin A treatment. Three of the four hospital-based studies (Barclay 1987; Coutsoudis 1991; Hussey 1990) which used the two-dose regimen demonstrated a protective effect on mortality. These studies were done under controlled conditions and their follow up was relatively brief. Only the Coutsoudis study indicated some long-term benefit of vitamin A as children were followed for six months; the outcomes used for this review were at the time of discharge from hospital. These factors affect the generalizability of the results to the general population of patients with measles.
An absence of vitamin A effect, or a smaller effect, in the community studies (Dollimore 1997; Rosales 1996) may be due to the study populations being healthier than the studies in hospitals. The community studies did not include children who were very sick as they were referred for hospitalization. The Rosales and Dollimore studies differed from the other studies in patient setting, follow up, disease severity, patient age, vitamin A preparation used and analytical approach. They looked at ambulatory participants who were followed up closely for one month with daily and weekly visits to urban health centres. This reflects the patient-care conditions under which the majority of measles cases are diagnosed and treated in low-income countries (Dollimore 1997; Rosales 1996).
Baseline differences and the presence of complications on admission
The demographic, nutritional, immunological and clinical status at baseline all affect the comparability between the vitamin A-treated and control groups (Coutsoudis 1991). Although all the studies reported the baseline nutritional status of the vitamin A-supplemented and placebo groups only, Barclay specified the nutritional status of the children who died; vitamin A recipients suffered lower mortality at every nutritional level.
In the Ogaro study (Ogaro 1993) 10 children were severely malnourished in the vitamin A-supplemented group and five children in the placebo group. This raises an issue about whether randomization balanced this important confounder. This could have been an important difference, possibly resulting in an inability to demonstrate a protective effect of vitamin A in the supplemented group. All the deaths in this study were due to pneumonia (five in the vitamin A group and three in the placebo group).
Five of the studies were carried out in Africa. The baseline prevalence of vitamin A deficiency and other baseline characteristics vary across countries and even within the same country, as in South Africa. The health services in the five areas of the included studies could be different and this could be one of the reasons, in addition to dose, that the studies showed different results.
Rosales suggested that as the population in his study was a healthier population than in previous studies this may explain an absence of, or smaller, vitamin A effect compared with that found in other studies (Rosales 1996).
In Hussey's study 64% of children had diarrhea and pneumonia on admission; while in Barclay's study pneumonia was the most frequent complication, affecting 85 children: 43% in the vitamin A group and 51% in the control group.
Most of the morbidity outcomes are either based on single or two studies, except for croup. As all studies did not report on all possible morbidity outcomes the conclusions we were able to draw about the effect of vitamin A on measles-related morbidity are limited.
There was a significant decrease in the incidence of croup with vitamin A supplementation while there was no significant reduction in the incidence of pneumonia, although a reduction was observed in the duration of diarrhea, pneumonia, fever, hospital stay and cough. Treatment of measles cases with vitamin A also has relevance to high-income countries as a reduction is seen in morbidity outcomes in Kawasaki's study. The Kawasaki (Kawasaki 1999) study reported no mortality and the morbidity outcomes were not pooled with those of the other studies as this study was from a developed country, that is Japan; it used only a single dose of 100,000 IU of vitamin A.
Limitations of this review
Nutritional status is an important predictor of vitamin A deficiency and mortality. The small number of studies and sample sizes have made it difficult to stratify or do a meta-regression. The subgroup analyses are very restricted as the same studies are represented in all of them. The apparent differences between trials may be related to the subgroup but could equally be confounded by some other aspect of trial design.
In these trials it was not always apparent as to which day after the onset of measles vitamin A was administered. Another limitation is that the follow-up period is not the same in all studies. It is assumed that all have been followed up until they were discharged from hospital. For the purposes of this review, the outcomes were taken at the time of discharge, hence it is not possible to make comparisons for delayed mortality across these studies.
It would have been useful to have the baseline incidence of measles in the study populations reported and if there were epidemics during the study period. The cases enrolled during a measles epidemic could vary in severity from measles cases at other times.
There was also a lack of reporting on the immunization status of children in the general population and in the study population, which was reported in only two studies (Dollimore 1997; Hussey 1990). The level of immunization would have had an impact on the severity of measles as it could reduce the intensity of exposure and hence the dose of the infecting virus (Hussey 1997). This would have had an impact on the severity of the disease as well as the severity of any epidemic. The severity of measles would be less in already vaccinated children (showing vaccine failure) and in areas where the immunization coverage was high.
Vitamin A is not only effective but also cost-saving. Hussey (Hussey 1990) demonstrated that the duration of hospital stay for children given vitamin A was decreased by an average of 4.7 days; by half a day in another study (Kawasaki 1999). The cost of a dose of vitamin A is around USD 0.02 (WHO 1998). At this cost ... "to achieve significant reductions in hospitalizations and costs in terms of mortality and long-term morbidity, vitamin A therapy for the management of measles is highly cost-effective" (Cervinskas 1996).
Until 1993, there were no reports of acute vitamin A toxicity in children with measles who took the WHO recommended dose as reported by the Committee on Infectious Diseases of the American Academy of Pediatrics (Pediatrics 1993). Even doses up to 400,000 IU have been reported to be relatively safe (Frieden 1992). None of the studies included in this review reported any adverse effects.
Headaches, loss of appetite, vomiting and bulging fontanelles (in infants) are some of the known adverse effects occasionally occurring with the administration of high doses of vitamin A. However, these symptoms are minor and transitory, with no known long-term effects and requiring no special treatment (WHO 1998). Under these circumstances it would appear that two doses of vitamin A are not too expensive, not likely to produce adverse effects and still have the capacity to reduce morbidity and mortality.
Overall completeness and applicability of evidence
In these trials there was a lack of information regarding the baseline vitamin status of these children, and also a lack of information about whether children had received previous vitamin A supplements prior to the onset of measles.
Not every study collected information on recovery from morbidity. We had some concerns about whether some trial authors collected data but later chose not to report these findings. As there were many outcomes reported by single studies there is the possibility that some effects would appear to be significant by chance alone.
Quality of the evidence
We have analyzed the risk of bias in all individual trials in detail in 'Risk of bias' tables. Overall, the quality of evidence for the use of vitamin A for measles in children can be considered as moderate.
Potential biases in the review process
Publication bias cannot reasonably be assessed in this review.
Agreements and disagreements with other studies or reviews
The conclusions of this review are in keeping with the previous three reviews (Beaton 1993; Fawzi 1993; Glasziou 1993), which were carried out at a time when only three trials (Barclay 1987; Coutsoudis 1991; Hussey 1990) were available. These are also the studies using two doses and showing a protective effect on measles mortality in the children treated with vitamin A. Later studies (Dollimore 1997; Kawasaki 1999; Ogaro 1993; Rosales 1996) used a single dose or more doses of oil-based vitamin A and did not show reduced measles mortality. Hence, authors of earlier reviews were not able to compare dosages in subgroup analyses. In addition, Fawzi's meta-analysis (Fawzi 1993) included Ellison's study of 1932 (Ellison 1932). Although it is a large study it has been included in this review only as part of the sensitivity analysis as it received a low quality score. It may also be worth mentioning that the objectives of those reviews were different from the objective of this review.
The findings of this review are consistent with one of the largest observational studies that reported on mortality as an outcome (Hussey 1997). A retrospective hospital record review of 1720 cases of measles, during 1985 to 1986, and 1989 to 1990, was carried out. There were 651 children in the latter time period who received two doses of vitamin A (200,000 IU) and had a shorter hospital stay, lower requirement for intensive care and lower death rate as compared to 1069 children during 1985 to 1986 who received a single dose of 3000 IU.
This review confirms that two doses of vitamin A are associated with a statistically significant reduction in the risk of overall mortality. The only conclusion that can be drawn with any degree of certainty is that high doses of oil- or water-based vitamin A are associated with greater reductions in mortality in children under the age of two years. It is possible that, in high doses, oil-based and water-based vitamin A have similar effects in children under the age of two years. Therefore, in this age group formulation did not make any difference. On the other hand, as studies that used two doses were also done in high case-fatality areas, there was no evidence to show that a single dose would not be effective as there were no studies using a single dose of oil-based vitamin A in these areas. Similarly, subgroup analyses by causes of morbidity and mortality by age group and formulation could not be done as the information was not available.