Papers

Meta-analysis of trials of prophylactic antibiotics for children with measles: inadequate evidence

BMJ 1997; 314 doi: https://doi.org/10.1136/bmj.314.7077.334 (Published 01 February 1997) Cite this as: BMJ 1997;314:334
  1. Frank Shann, professor of critical care medicinea
  1. a Intensive Care Unit, Royal Children's Hospital, Parkville, Victoria 3052, Australia
  • Accepted 15 November 1996

Abstract

Objective: To assess whether antibiotics should be given to all children with measles in communities with a high case fatality rate.

Design: Meta-analysis of randomised controlled trials that compared routine antibiotic prophylaxis with no antibiotic treatment or selective treatment of pneumonia or sepsis.

Subjects: Six trials of children admitted to hospital with measles: five in Glasgow, London, or New York between 1939 and 1954; and one in India in 1967.

Main outcome measures: Incidence of pneumonia or sepsis, and mortality.

Results: All but one of the trials were unblinded, and randomisation was either not described or was by alternate allocation. In four studies, the incidence of pneumonia or sepsis in the control group was similar to that in the antibiotic prophylaxis group; in the other two studies, the incidence of pneumonia or sepsis was unusually high in the control group so these children had a higher complication rate than the antibiotic group. Four of the 764 children given antibiotics died compared with one of the 637 controls (exact odds ratio 4.0, mid-P corrected 95% confidence interval 0.5 to 101.6).

Conclusion: The quality of the trials reviewed was poor, and they provide weak evidence for giving antibiotics to all children with measles. Available evidence suggests that, when mortality from measles is high, all children with measles should be treated with vitamin A but antibiotics should be given only if a child has clinical signs of pneumonia or other evidence of sepsis.

Key messages

  • Since many deaths from measles in young children are caused by bacterial pneumonia it might be sensible to give antibiotics to all children with measles where there is high case mortality, but there has been no systematic evaluation of such a policy

  • Six randomised controlled trials of prophylactic antibiotics in children with measles were found

  • All trials were poorly designed and preceded the introduction of vitamin A for measles

  • Prophylactic antibiotics had little or no effect on mortality from measles

  • In communities with mortality from measles of 1% or more, all children with measles should be treated with vitamin A but antibiotics should be given only if a child has clinical signs of pneumonia or sepsis

Introduction

Measles causes more than a million deaths a year, and most of these deaths are from pneumonia in children less than 5 years old.1 There is evidence that pneumonia in children with measles is often caused by bacterial infection. Histological features of bacterial pneumonia were found at necropsy in nine of 26 children in Uganda,2 and five of 21 children in South Africa.3 Bacteria were isolated from lung aspirates in 71 of 75 children studied in London in 19314 and, in more recent studies, from 15 of 22 children in South Africa,5 two of 21 children in Colombia,6 and 20 of 56 children in Nigeria.7 In adults with measles and pneumonia in the United States, bacteria were isolated from transtracheal aspirate in 10 of 16 patients in one study8 and 35 of 106 patients in another.9

The evidence that pneumonia in children with measles is often caused by bacteria suggests that, in countries where there is a high case mortality, it might be sensible to give antibiotics to all children with measles. However, there has been no systematic evaluation of the evidence for such a policy, and there is good evidence that routine antibiotic prophylaxis does not prevent pneumonia developing in children with upper respiratory tract infections caused by other viruses.10 To assess the strength of the evidence for giving prophylactic antibiotics to all children with measles, I performed a meta-analysis of randomised controlled trials in which routine antibiotic prophylaxis was compared with no treatment or selective treatment for pneumonia or sepsis.

Methods

I performed a Medline search for the years 1966-95, searching on measles plus either antibiotics, penicillins, sulphonamides, prospective studies, or randomised controlled trials. I also used the results of my previous study of pneumonia in children which included a hand search of every journal in the University of Melbourne medical library for the period 1935-46 inclusive.11

For each study, I sought information about the selection of patients; randomisation procedures; exclusions after randomisation; and the number of patients randomised, the number who developed pneumonia or sepsis, and the number who died. In my analyses of the development of pneumonia or sepsis and of mortality nearly half the cells contained five or fewer items, so I performed the meta-analyses with StatXact.12

Result

Studies

I identified six randomised controlled trials of the use of prophylactic antibiotics in children with measles (table 1).13 14 15 16 17 18 Five of the trials were conducted in Glasgow, London, or New York between 1939 and 1954,13 14 15 16 17 a time when measles was often complicated by bronchopneumonia but controlled trials were not well designed; the sixth study was conducted in India in the 1960s.18 Apart from two studies in which the method of randomisation was not specified,15 18 alternate children were allocated to either an antibiotic or a control group when they were admitted to the trial; in four studies antibiotics were given to all children who had pneumonia at the time of admission and to any child in the control group who developed pneumonia or sepsis,15 16 17 18 and the other two studies did not give any antibiotics to children in the control group.13 14 All but one of the studies were unblinded,13 14 15 16 17 and only one paper mentioned withdrawals from the trial.13 None of the papers provided information about antibiotic treatment given before randomisation.

View this table:
Table 2

Number (percentage) of children developing pneumonia or sepsis in randomised controlled trials included in meta-analysis of antibiotic prophylaxis for children with measles

View this table:

Three published studies were excluded from this analysis because they were not randomised.19 20 21 Thompson reported that bronchopneumonia developed in 1.7% of 352 children treated with an antibiotic, and in 4.8% of 762 controls.19 On the other hand, Weinstein found pneumonia at the time of admission to hospital in 21.5% of 130 children who had been treated with an antibiotic, and 8.1% of 298 children who had not had an antibiotic.20 In Senegal the case mortality from measles fell substantially after an increase in measles immunisation and administration of co-trimoxazole to all children less than 3 years old who had measles21; however, immunisation has been shown to reduce the case mortality from measles,22 23 and the co-trimoxazole may have had little or no effect.

Development of pneumonia or sepsis

In my analysis of the development of pneumonia or sepsis after admission (table 2), the exact test for homogeneity showed that there was no common odds ratio across the six 2x2 (P<0.00005), so I did not report the odds ratio. The scanty details provided in the papers and the small sample sizes meant that the cause of the lack of homogeneity could not be determined. It was possible to reject the null hypothesis that all the 2x2 in table 2 had an odds ratio of 1.0 (exact two sided test P=0.0004); however, the P value was 0.833 when the two studies with an unusually high mortality in the control group17 18 were removed.

In the first four studies13 14 15 16 the incidence of pneumonia or sepsis in the control group was similar to that in the antibiotic prophylaxis group (11/462 (2%) v 12/447 (3%)); in the other two studies,17 18 where there was an unusually high incidence of pneumonia or sepsis in the control group, there were more complications in the control group than in the antibiotic group (23/161 (14%) v 2/234 (1%)).

Mortality

There were four deaths in the 764 children given prophylactic antibiotics, and only one in the 637 children in the control group (table 3). The exact test for homogeneity gave a P value of 1.00, and the exact common odds ratio was 4.0 (mid-P corrected 95% confidence interval 0.5 to 101.6). There is, therefore, no evidence that mortality was lower when antibiotics were given routinely rather than selectively or not at all.

Table 3

Number (percentage) of children who died in randomised controlled trials included in meta-analysis of antibioticprophylaxis for children with measles

View this table:

Discussion

The main limitations of this analysis were that the six controlled trials included a total of only 1304 children and all six trials were poorly designed–randomisation was not described or was by alternate allocation, all but one were unblinded, and little information was provided about withdrawals. Only one of the trials13 had a mortality of 1% or more, although two others14 15 were in communities where the mortality had been more than 1% in the immediately preceding years, when antibiotics were not available.

Because of the poor quality of these studies, any inferences must be tentative. The studies suggest that routine antibiotic treatment might reduce the risk of developing pneumonia or sepsis (P=0.0004), but the effect was small and disappeared when the two studies with unusually high complication rates in the control group were excluded (table 2). These findings are consistent with the evidence that routine antibiotic treatment of non-measles upper respiratory tract infection does not prevent pneumonia.10

The studies reviewed here provide no evidence that routine antibiotic treatment resulted in a lower mortality (four deaths in 764 children) than no antibiotic treatment (one death in 232 children) or selective treatment when complications developed (no deaths in 405 children) (table 3). However, the small number of deaths means that the possibility of a reduced mortality from routine prophylaxis cannot be excluded. Some of the children who died may have already had pneumonia or sepsis at the time they were admitted to the trials. Antibiotic treatment of cases with pneumonia or sepsis seemed to have a considerable effect on mortality: for example, at the Kingston Avenue Hospital, New York, 2.63% of 3611 children with measles died in 1935-40, before antibiotics were used, compared with only 0.74% of 1213 children in 1941, when sulfathiazole was first used.15

Recommendations

Mortality from measles is said to be high if it is 1% or more.24 How should children with measles be managed in communities with a high mortality? All such children should be treated with high doses of vitamin A, which substantially reduces mortality from measles.24 25 Children who have clinical signs of pneumonia (cyanosis, inability to feed, chest indrawing, or tachypnoea) or other evidence of sepsis should be given antibiotics,26 but should antibiotics be given to children with measles who have no clinical signs of pneumonia or sepsis? The controlled trials of prophylactic antibiotics for measles provide only weak evidence that they reduce the incidence of pneumonia and no evidence that they reduce mortality. In addition, the trials were performed before the introduction of vitamin A treatment for measles and before the development of detailed guidelines for the early diagnosis of pneumonia in children. The routine administration of prophylactic antibiotics to all children with measles would be expensive, and it would encourage the development of antibiotic resistance.

To prevent unnecessary admissions, conserve scarce resources and reduce cross infection, it is important that only high risk cases of measles are admitted to hospital in developing countries.27 In these circumstances a high proportion of children admitted to hospital will have pneumonia or sepsis and will therefore need treatment with an antibiotic.

The poor quality of the trials reviewed here suggests a need for further studies. However, a controlled trial of routine antibiotic prophylaxis would need 3000 patients in each group to detect a halving in mortality from 2% to 1% with a two sided α error of 0.05 and a ß error of 0.2, and it would be a major exercise to mount such a large study in a developing country. Given the pressing need for large controlled trials of vaccines and trials of the treatment of established pneumonia in the presence of antibiotic resistant organisms, a controlled trial of prophylactic antibiotics for measles has a low priority.

Acknowledgments

Funding: None.

Conflict of interest: None.

References

  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
View Abstract