Non-specific beneficial effect of measles immunisation: analysis of mortality studies from developing countriesBMJ 1995; 311 doi: http://dx.doi.org/10.1136/bmj.311.7003.481 (Published 19 August 1995) Cite this as: BMJ 1995;311:481
- Peter Aaby, senior researchera,
- Badara Samb, field physicianb,
- Francois Simondon, project directorb,
- Awa Marie Coll Seck, professorc,
- Kim Knudsen, senior statisticiana,
- Hilton Whittle, deputy directord
- a Epidemiology Research Unit, Danish Epidemiology Science Centre, Statens Seruminstitut, Artillerivej 5, 2300 Copenhagen, Denmark,
- bORSTOM, Dakar, Senegal
- cUniversity Cheikh Anta Diop, Dakar, Senegal,
- d Medical Research Council Laboratories, Banjul, Gambia
- Correspondence to: Dr Aaby.
- Accepted 19 August 1995
Objective: To examine whether the reduction in mortality after standard titre measles immunisation in developing countries can be explained simply by the prevention of acute measles and its long term consequences.
Design: An analysis of all studies comparing mortality of unimmunised children and children immunised with standard titre measles vaccine in developing countries.
Studies: 10 cohort and two case-control studies from Bangladesh, Benin, Burundi, Guinea-Bissau, Haiti, Senegal, and Zaire.
Main outcome measures: Protective efficacy of standard titre measles immunisation against all cause mortality. Extent to which difference in mortality between immunised and unimmunised children could be explained by prevention of measles disease.
Results: Protective efficacy against death after measles immunisation ranged from 30% to 86%. Efficacy was highest in the studies with short follow up and when children were immunised in infancy (range 44-100%). Vaccine efficacy against death was much greater than the proportion of deaths attributed to acute measles disease. In four studies from Guinea-Bissau, Senegal, and Burundi vaccine efficacy against death remained almost unchanged when cases of measles were excluded from the analysis. Diphtheria-tetanus-pertussis and polio vaccinations were not associated with reduction in mortality.
Conclusion: These observations suggest that standard titre measles vaccine may confer a beneficial effect which is unrelated to the specific protection against measles disease.
In 10 cohort studies measles efficacy against death was in the range of 30-86%
The specific prevention of the acute and long term consequences of measles disease does not explain the reduction in mortality among immunised children
In three studies diphtheria-tetanus-pertussis and polio vaccines were not associated with similar reductions in mortality, making it unlikely that selection bias can explain the impact of measles immunisation
Standard titre measles vaccine seems to be associated with a non-specific, beneficial effect which may have important implications for the planning of immunisation programmes
Evaluations of immunisation programmes are usually based on the assumption that vaccines have an impact only against specific diseases. This assumption may not be correct for measles vaccine. Recent studies indicate that vaccines may have important non-specific effects as girls receiving high titre measles vaccines were found to have reduced long term survival compared with recipients of standard titre vaccines.1 2 3 On the other hand, studies of standard titre measles vaccine have reported a greater than expected reduction in mortality in areas with high mortality.4 5 6 As these observations suggest that measles immunisation may have a non-specific, beneficial effect5 we reviewed mortality studies of unvaccinated and vaccinated children and examined whether the reduction in mortality after measles immunisation is due only to the specific prevention of acute measles disease and its long term consequences. If measles vaccines have non-specific, beneficial effects the age at immunisation and the number of doses of vaccines should be reconsidered. Furthermore, new measles vaccines would have to be evaluated for their impact on survival before being introduced, and immunisation would have to continue after possible eradication of measles unless the same beneficial effects could be produced through other means.
Studies and methods STUDIES OF STANDARD TITRE MEASLES VACCINE
We reviewed Index Medicus from 1970 onwards for studies dealing with mortality after standard titre measles vaccination. Table I shows the available studies with information on mortality among immunised and unimmunised children. We found 10 follow up studies and two case-control studies which had examined the impact of Schwarz standard titre measles vaccine.
STUDIES OF DIPHTHERIA-TETANUS-PERTUSSIS AND POLIO IMMUNISATION
Reduced mortality among recipients of standard titre measles vaccine compared with unimmunised children could be due to a selection bias between those children who attended and those who did not attend clinics for measles vaccination. We therefore examined whether diphtheria-tetanus-pertussis and polio vaccination was associated with a similar reduction in the areas where measles vaccine had also been examined. Attendance for diphtheria-tetanus-pertussis and polio vaccinations is probably associated with attendance for later measles immunisation. Therefore, any separate impact of these vaccines has to be examined at ages before measles immunisation. The only published study of this effect was a case-control study from Benin.16 Relevant data, however, were available from both Senegal and Guinea-Bissau.
We examined the impact of diphtheria-tetanus-pertussis and polio vaccines on mortality in children between 5 and 10 months of age in Niakhar, Senegal.2 12 At 5 months children were called for immunisation and some attended and received diphtheria-tetanus-pertussis, inactivated polio vaccine, and placebo for measles vaccine whereas others did not attend. At 10 months of age the children were called again for measles immunisation. The estimate of mortality ratio between 5 and 10 months was adjusted for previous immunisations at 3 months of age.
In Guinea-Bissau we used data from a national cluster sample of 10000 women of fertile age and their prospectively registered pregnancies (authors' unpublished data). Women of fertile age and their children were visited about every six months. In the present analysis we included only children whose immunisation card was seen and children who were assumed to be unvaccinated because they had no card. Children aged 2-3 months when first seen should have been immunised with diphtheria-tetanus-pertussis and oral polio vaccines and within six months of follow up they would not have received measles vaccine. Some children may have received other diphtheria-tetanus-pertussis and oral polio vaccines during follow up, but it was not possible to get full immunisation information for children who had died, moved, or were absent at the re-examination. To examine whether any vaccine is a marker for better survival we compared mortality of children aged 2-3 months during six months of follow up according to their immunisation status when first seen.
DEFINITIONS AND STATISTICAL METHODS
We have emphasised the crude estimates of mortality differences based on deaths by person years at risk available for all the follow up studies, but available multivariate estimates adjusted for significant background factors have also been noted in table II. Vaccine efficacy against death (VED) was calculated as one minus the mortality rate ratio between immunised and unimmunised children. We tested the homogeneity of the estimates of vaccine efficacy against death—that is, the hypothesis of no interaction between study and the size of the vaccine effect.17 The Mantel-Haenszel estimator was used to combine results from different subgroups.
Death from acute measles is usually defined as any death within one month8 or six weeks18 of a measles rash. In the present analyses we used the definition used by the study in question. Mortality after measles was considered as any death after the acute phase of measles, irrespective of whether it could be directly linked to measles disease. The possible impact of immunisation beyond the prevention of measles disease was assessed by comparing the mortality of immunised, uninfected children and unimmunised, uninfected children. This was possible in four studies (Guinea-Bissau III and IV, Senegal II and Burundi) by censoring follow up at the time of measles disease, thus excluding both death after acute measles and deaths after measles.
REDUCTION IN CHILDHOOD MORTALITY AFTER STANDARD TITRE MEASLES
Table II shows that in all 10 follow up studies the impact on mortality after standard measles immunisation was large, showing reductions in the range of 30-86%. The two case-control studies suggested similar reductions in mortality. Crude and adjusted estimates were virtually identical. All follow up studies showed large reductions, but the estimates of vaccine efficacy against death were heterogeneous (test for homogeneity, χ2=26.3; df=9; P=0.002; figure).
The follow up studies were of two kinds. The first kind compared attenders and non-attenders within the same community; vaccine efficacy against death was in the range of 38-86%. In studies comparing immunised and unimmunised children from different communities estimates of vaccine efficacy against death were less heterogeneous, showing estimates in the range of 30-67%. Other forms of heterogeneity, however, may have been more important for the variation in estimates in table II. The impact tended to be greatest in the studies when children were immunised early5 and which had a short follow up. For example, in seven studies from Zaire, Guinea-Bissau (I-IV), and Senegal (I-II) with further data available vaccine efficacy against death was higher, being in the range of 44-100%, when the analysis was limited to one year of follow up for children immunised in infancy than in the residual part of these studies (data available on request). In the Bangladesh II study, in which the mortality data were presented in three monthly intervals,15 vaccine efficacy against death was significantly greater in the first 24 months after immunisation (48%; 95% confidence interval 37% to 57%) than in the last 21 months of the study (6%; −46% to 40%) χ2=5.75; df=1; P=0.016). There were similar tendencies in the studies from Zaire, Senegal II, and Guinea-Bissau IV (data available on request).
STANDARD TITRE MEASLES VACCINE: PREVENTION OF ACUTE AND LONG TERM CONSEQUENCES OF MEASLES
As indicated in table II all studies found the reduction in mortality after measles immunisation to be much larger than the proportion of deaths attributed to acute measles disease. It has therefore been speculated that the prevention of delayed deaths from measles could explain the reduction.6 This could be tested by comparing mortality of unimmunised and immunised children after the exclusion of all cases of measles. If the impact of vaccine was related only to the specific prevention of the acute and long term consequences of measles disease there should be no difference in mortality according to immunisation status among uninfected children. This, however, was not the case in any of the studies (table III). In the three larger studies—Guinea-Bissau IV, Senegal II, and Burundi—there was no change in vaccine efficacy against death after exclusion of all cases of measles. Hence, in these studies the prevention of measles contributed little to the reduction in mortality associated with immunisation.
DIPHTHERIA-TETANUS-PERTUSSIS AND POLIO IMMUNISATIONS: NO ASSOCIATION WITH REDUCED CHILDHOOD MORTALITY
We also examined the impact of diphtheria-tetanus-pertussis and polio vaccines in areas where measles immunisation had been studied. In the case-control study from Benin recipients of one dose of diphtheria-tetanus-pertussis and oral polio tended to have higher mortality than unimmunised children (relative risk=2.20; 95% confidence interval 0.93 to 5.22).16 In the vaccine trial from Senegal II (table IV) the 638 children attending at 5 months and receiving diphtheria-tetanus-pertussis and inactivated polio vaccine (and placebo for measles vaccine) had slightly but not significantly higher mortality between 5 and 10 months of age than the 607 children not attending immunisation at 5 months (Mantel-Haenszel, mortality ratio=1.60; 0.76 to 3.37). In the cluster cohort study of 10000 women of fertile age and their children in Guinea-Bissau 488 children were 2-3 months old when first seen. During six months of follow up mortality was 4% (9/245) for children who had already received diphtheria-tetanus-pertussis and oral polio vaccines at least once and 3% (8/243) for children who had not received these vaccines.
In our analysis of studies on the protective efficacy against death of standard measles immunisation we found a reduction in mortality in the range of 30-86%. A major reduction in mortality after measles immunisation is also supported by a few studies comparing mortality rates before and after the introduction of measles vaccination.19 20 Though estimates were heterogeneous the reduction in mortality was considerably larger in all studies than the share of deaths attributed to acute measles disease in the same areas (table II). Surprisingly, the protective efficacy of measles vaccine was virtually unchanged when follow up was discontinued at the date of measles disease, suggesting that the reduction in mortality after measles immunisation may have little to do with the specific prevention of measles. Subclinical measles infection is rare after the age of measles immunisation,21 and it seems therefore unlikely that undetected measles infection is a major cause of higher mortality in the unimmunised group, particularly because clinical measles explained little of the difference in mortality. Several other observations also support the possibility that measles vaccine has non-specific effects. Contrary to expectations, several studies indicated that measles immunisation is particularly effective when given early in life.4 5 7 9 16 Furthermore, the reduction in mortality may be the greatest during the first year after immunisation as a higher vaccine efficacy was observed when the follow up period was limited to one year.7 10 12 15 Though few studies have reported data y6 by sex it seems that standard vaccine may be more beneficial for girls than for boys.12 19 22
Double blind placebo trials of standard titre measles vaccine on mortality in developing countries have not been performed, and as differences in mortality were not explained by prevention of measles the difference between immunised and unimmunised children could reflect an association between measles immunisation and access to other health interventions or a selection bias. Most studies (Bangladesh I and II, Guinea-Bissau I, II, and III, Senegal I, Zaire) excluded an association with other health interventions because measles immunisation was the only intervention available or the only intervention which differed between the areas. Most studies (Bangladesh I and II, Guinea-Bissau II, III, and IV, Haiti, Zaire) tried to exclude the possibility that selection bias was the major cause of differences in mortality by using multivariate analysis to adjust for important determinants of mortality (table II), by showing no difference between the groups before the introduction of measles vaccine, or by comparing those who did not seroconvert because they had received a placebo and children who had received an effective vaccine.
A bias due to publication of only those studies with significant results seems unlikely as a strong effect of measles immunisation has been reported from almost all the longitudinal research on measles or measles immunisation.4 6 7 8 9 10 11 12 14 15 18 19 23 Though the estimate for vaccine efficacy against death was slightly lower (30-67%) for the more satisfactory studies comparing immunised and unimmunised areas than for the other studies (38-86%) comparing immunised and unimmunised children from the same area, all studies documented the same unexplained reduction in mortality.
If a systematic selection bias between attenders and non-attenders was the main cause of the clear impact of measles immunisation a similar difference in mortality could be expected between recipients and nonrecipients of diphtheria-tetanus-pertussis and polio vaccines, particularly as these vaccines are given early in life when mortality is high. In the three areas with relevant data, there was no indication that immunisation with these vaccines were also associated with reduced mortality.
The observation that exclusion of cases of measles had little effect on the vaccine efficacy against death contradicts previous studies that suggest that measles is associated with a significant long term excess mortality.4 24 Previous studies compared mortality after measles with mortality in immunised controls, however, rather than with unimmunised children who are the appropriate controls if measles immunisation has non-specific effects. For example, children who had survived the acute phase of measles in Guinea-Bissau were found to have significantly higher mortality than community controls who had received measles vaccine (mortality ratio 4.18; 1.13 to 15.43).4 Compared with unimmunised controls, however, children who survived the acute phase did have slightly lower mortality (0.45; 0.14 to 1.43). More recent analyses of the long term effect of measles disease in Guinea-Bissau, Senegal, Bangladesh (authors' unpublished observation), and Burundi13 indicate that children who survive acute measles have the same or significantly lower mortality than non-infected unimmunised children. Hence, acute mortality may partly be compensated by lower subsequent mortality, and the total mortality impact of measles in the unimmunised group may be limited.
If protection against measles disease does not explain the impact of measles immunisation on child survival the simplest explanation would seem to be that measles vaccine activates the immune system in a non-specific way providing protection against other infections. Studies of immune responses to measles infection have mainly focused on immunological abnormalities possibly explaining the expected immunosuppression and increased susceptibility to other infections leading to complications and death.25 Immunological stimulation by measles disease and immunisation, however, may also protect against other infections.26 27 For example, measles immunisation reduces the incidence of diarrhoea (authors' unpublished observation) and may prevent subsequent immunisation with vaccinia.26
The hypothesis of a non-specific beneficial effect of measles vaccine has important practical and theoretical implications. If new vaccines do not provide similar non-specific effects, new measles vaccines capable of immunising in the presence of maternal antibodies28 may end up being associated with lower survival than standard titre measles vaccine. The available data indicate that child survival might benefit from standard titre measles immunisation before 9 months of age and possibly also from repeated doses of the vaccine.5 Further studies are obviously needed to explain the biological basis and to determine the magnitude of the non-specific effects. Such studies may be conducted within two dose trials or studies of the impact of reimmunisations. Should the hypothesis be correct measles immunisation may have to be continued even when measles infection has been eradicated.
Funding Danish Council for Development Research, Danish Council for Medical Research, Copenhagen, Denmark; Task Force for Child Survival and Development, Atlanta, United States; Expanded Programme on Immunization, WHO, Geneva; Science and Technology for Development Programme of the European Community, Bruxelles, Belgium (TS3*-CT91-0002); UR Population et Sante, ORSTOM, Dakar, Senegal; and UNICEF, Guinea-Bissau.
Conflict of interest None.