Non-specific effects of measles, mumps, and rubella (MMR) vaccination in high income setting: population based cohort study in the Netherlands
BMJ 2017; 358 doi: https://doi.org/10.1136/bmj.j3862 (Published 30 August 2017) Cite this as: BMJ 2017;358:j3862
All rapid responses
In their rapid response, Benn et al. discussed the sequence of vaccination to assess non-specific effects of vaccines. Specifically, they discussed the results of our analysis on the admission risk for children receiving DTaP-IPV-Hib-4+PCV after MMR+MenC. This is a so-called ‘reversed schedule’ because the recommended schedule in the Netherlands is to receive the fourth DTaP-IPV-Hib+PCV vaccine at 11 months of age and the MMR+MenC vaccine at 14 months of age. Reviewer 2 requested to do this analysis and therefore we provided the results in the response to the reviewers (the HR of 0.85 (95% CI: 0.70-1.03) given in the response to reviewers is for all admissions, the estimate for >1 day admissions was 0.93 (95% CI: 0.67-1.31)). However, we decided not to include the estimate in the manuscript because it is a very specific and very small group of children receiving this reversed schedule. Only 3.3% of the total study population (44,653/1,356,926) was included in this analysis. In this group, many more children had parents who were born outside the Netherlands (40% vs 11% in the ‘recommended schedule’ population) and parental education level was lower. Because this is such a specific and small group, we think we should be hesitant in interpreting this estimate in the context of non-specific effects.
Benn et al. subsequently compared the estimate of the reversed schedule with the estimate of the recommended schedule in their response to estimate non-specific effects. We think we should be careful with comparing estimates from different populations, because it requires an assumption that biases, including the healthy vaccinee bias, are the same in these populations. Because the populations are different in terms of e.g. age, parental country of birth, parental education level and possibly other unknown factors, the occurrence or magnitude of healthy vaccinee bias may also be different. In general, we think that these comparisons do not help to disentangle possible non-specific effects and bias.
Benn et al. ask us to report the results of the reversed schedule also by type of infection. However, as discussed above we think the estimates of the analyses in this specific reversed schedule group should be interpreted with caution. Moreover, validity of comparisons between the results of the analyses on the reversed schedule and the recommended schedule rely on assumptions we cannot prove.
Although we encourage searching for smart designs or methods to try to disentangle non-specific effects and (healthy vaccinee) bias, we have to acknowledge that it is problematic to study non-specific effects in observational studies, and that, in the end, randomized controlled trials are needed to obtain a definitive answer.
Competing interests: No competing interests
The sequence of vaccinations to test the non-specific effects of vaccines
There is growing evidence that vaccines may modify susceptibility to unrelated infections with implications for hospital admissions patterns in high-income countries (1,2) and mortality levels in low-income countries (3). Ideally, such potential non-specific effects of vaccines (NSEs) should be tested in randomised clinical trials (RCTs). However, that is unlikely to happen due to the ethical challenges related to randomising children to recommended vaccines. Hence, we need other ways of testing whether vaccines are likely to have important NSEs.
The disease-specific effects of vaccines are likely to be the same irrespective of the sequence in which the vaccines were provided; for example, children vaccinated with diphtheria-tetanus-pertussis (DTP3) and measles-mumps-rubella (MMR) vaccines are likely to be protected against pertussis and measles infections irrespective of whether they got MMR before or after DTP3. However, many observational studies (3) and RCTs (4) have shown that the sequence may matter for the NSEs of vaccines, the most recent vaccination having the strongest effect.
Studies from low-income countries suggest that it is better to have measles vaccine (MV) after DTP3 rather than to have DTP3 after MV (3,4). Similarly, in the Danish register based studies of hospital admission (1), MMR was associated with 14% (12-16%) lower risk of admission than children still having DTaP-IPV-Hib-3 as their most recent vaccination, whereas getting the vaccines in the opposite sequence, i.e. DTaP-IPV-Hib3 after MMR, was associated with a 62% (28-105%) increase in the risk of admission. The increase was particularly strong for upper respiratory infections (89% (31-74%)) and lower respiratory infections (64% (6-154%)).
The recent paper on the potential NSEs of vaccines in the Netherlands did not report the admission risk for children receiving DTaP-IPV-Hib-4+PCV after MMR+MenC (5). However, in the response to reviewers (available at the BMJ website), Tielemans et al. mention that the adjusted HR was 0.85 (0.70-1.03) for DTaP-IPV-Hib-4+PCV (after MMR+MenC) compared with having MMR+MenC (after DTaP-IPV-Hib-3+PCV) as last vaccination. It is not indicated whether this estimate refer exclusively to the >1 day admissions or to all admissions. In the main paper, compared with children who still had DTaP-IPV-Hib-4+PCV as their last vaccination, MMR+MenC (after DTaP-IPV-Hib-4+PCV) was associated with HRs of 0.62 (0.57-0.67) for >1 day admissions and 0.40 (0.38-0.41) for all admissions (5).
This would suggest that the differential effect of having DTaP-IPV-Hib-4+PCV after MMR+MenC rather than MMR+MenC after DTaP-IPV-Hib-4+PCV as the last vaccine was a 37% (11-69%) (0.85/0.62=1.37 (1.11-1.69)) increase in the relative risk of admission if the 0.85 estimate referred to >1 day admissions. [The differential effect would be 113% (75-159%) (0.85/0.40=2.13 (1.75 -2.59)) increase in the relative risk of admission if the 0.85 estimate referred to all admissions].
Hence, the data suggest that DTaP-IPV-Hib-4+PCV after MMR+MenC is associated with an increase in relative risk of admissions compared to MMR+MenC after DTaP-IPV-Hib-4+PCV. This is similar to the pattern in Denmark (1) and low-income countries (3,6). Unless it can be shown that healthy vaccinee bias works differently in the two situations, the pattern is likely to reflect that MMR+MenC and DTaP-IPV-Hib-4+PCV are associated with completely different NSEs.
To test our observation that the negative effect of the reverse sequence was strongest for respiratory infections, we would appreciate if Tielemans et al. would report the results for DTaP-IPV-Hib-4+PCV after MMR+MenC also for the four admission groups used in the main paper (lower respiratory infections, upper respiratory infections, gastrointestinal infections, other infections) (5).
The sequence of vaccinations have received relatively little attention (3,4,6). This is unfortunate since analysing the sequence may help to assess the relative importance of NSEs of different vaccines, even in situations with severe healthy vaccinee bias.
So far the results have been consistent in indicating that it is better to receive a live vaccine after a non-live vaccine than a non-live after a live vaccine (1,3,4,6). Hence, it could have major effects on mortality and morbidity and health care costs if immunisation programmes implemented a “live-vaccine-last” policy.
References
1. Sørup S, Benn CS, Poulsen A, Krause TG, Aaby P, Ravn H. Live vaccine against measles, mumps, and rubella and the risk of hospital admissions for nontargeted infections. JAMA 2014;311:826-35
2. Bardenheier BH, McNeil MM, Wodi AP, McNicholl J, DeStefano F. Risk of nontargeted infectious disease hospitalizations among U.S. children following inactivated and live vaccines, 2005-2014. Clin Inf Dis 2017;epub
3. Higgins JPT, Soares-Weiser K, López-López JA, Kakourou A, Chaplin K, Christensen H, Martin NK, Sterne JAC, Reingold AL. Association of BCG, DTP, and measles containing vaccines with childhood mortality: systematic review. BMJ 2016;355:i5170
4. Aaby P, Ravn H, Benn CS, Rodrigues A, Samb B, Ibrahim SA, Libman MD, Whittle HC. Randomized Trials Comparing Inactivated Vaccine After Medium- or High-titer Measles Vaccine With Standard Titer Measles Vaccine After Inactivated Vaccine: A Meta-analysis. Pediatr Infect Dis J. 2016 Nov;35(11):1232-1241
5. Tielemans SMAJ, de Melker HE, Hahné SJM, Boef AGC, van der Klis FRM, Sanders EAM, van der Sande MAB, Knol MJ. Non-specific effects of measles, mumps, and rubella (MMR) vaccination in high income setting: population based cohort study in the Netherlands. BMJ 2017;358:j3862
6. Welaga P, Oduro A, Debpuur C, Aaby P, Ravn H, Andersen A, Binka F, Hodgson A. Fewer out-of-sequence vaccinations and reduction of child mortality in Northern Ghana. Vaccine 2017;35:2496-2503
Competing interests: No competing interests
Beneficial non-specific reduction in hospital admissions for respiratory infections following MMR and MenC vaccinations in the Netherlands
We appreciate that Tielemans et al.(1), using Dutch population data on vaccinations and hospital admission for infectious diseases, tested our findings of non-specific effects (NSEs) of the live attenuated measles-mumps-rubella (MMR) vaccine in Denmark(2). We found the transition from the third DTaP-IPV-Hib (DTP3) to MMR to be associated with a hazards ratio (HR) of 0.86 (95%CI=0.84-0.88). The Dutch study finds much stronger beneficial effect estimates for the transition from the fourth DTaP-IPV-Hib-PCV (DTP4) to MMR+MenC (HR for all admissions of 0.40 (0.38-0.41); for >1 day admissions 0.62 (0.57-0.67)), but also strong beneficial effect estimates for the previous transition from DTP3-to-DTP4 (HR all admissions 0.48 (0.46-0.51); >1 day 0.69 (0.63-0.76)). The authors conclude that healthy vaccinee bias at least partly explains the lower rates of admission after MMR and add that though NSEs cannot be excluded, NSEs cannot be distinguished from bias. The authors therefore emphasise the importance of caution in the interpretation of observational studies on the NSEs of vaccines.
There are several important contextual differences between the Danish and the Dutch studies. For instance, Denmark provides MMR at 15 months, whereas the Netherlands provides MMR+MenC vaccines at 14 months, and the NSEs of a live vaccine vs. co-administration of a live vaccine and a non-live vaccine have been shown to differ(3).
Most important in relation to healthy vaccinee bias is the difference in the organisation of vaccination. In the Netherlands, all children have pre-scheduled vaccination appointments, so 99.6% of all children had received MMR+MenC vaccine by 24 months of age. Judged by Figure 2(1), it was the sick children, who were delayed, since the admission rate increased in the unvaccinated children after 14 months. If almost all are vaccinated except sick children, this creates a strong healthy vaccinee bias, or maybe rather “unhealthy non-vaccinee bias”. In Denmark, parents have to schedule the vaccination appointments themselves; less than 90% had received MMR by 24 months; therefore the group of delayed children is more diverse and the admission rate did not increase in the unvaccinated children after 15 months(2).
It could be discussed whether it is meaningful to compare vaccinated versus unvaccinated children in settings like the Dutch, with severe healthy vaccinee bias. Tielemans et al. propose one way to control for healthy vaccinee bias: if we assume that the HR associated with the transition from DTP3–to-DTP4 is due to healthy vaccinee bias, then the relative difference in the HRs for DTP3-to-DTP4 and DTP4-to-MMR+MenC could be ascribed to the NSEs of MMR+MenC(1). The authors do not present the calculation, but if we use the HR for all admissions for the transition from DTP3-to-DTP4 as an indication of the “baseline” healthy vaccinee bias, then an adjusted HR for all admissions for the transition from DTP4-to-MMR+MenC can be estimated as 0.40/0.48=0.83 (95% CI=0.78-0.89). The correction of a HR for a vaccine by adjustment for the HR of another vaccine represents one way to deal with healthy vaccinee bias. It cannot be excluded that healthy vaccinee bias could be different in different age groups and the true estimate of the NSEs of MMR+MenC might well be larger or smaller than 17% (11-22%).
Are there other ways to control for healthy vaccinee biases? We have previously used triangulation methods to examine whether NSEs are the most logical explanation for all the data available. In the Danish study, we showed that DTP administered after MMR was associated with increased risk of admission, a tendency which cannot be explained by healthy vaccinee bias. Furthermore, both observational studies and RCTs(4) from low-income countries have suggested that measles vaccine is associated particularly with lower risk of respiratory infections. We therefore tested whether the reduction in admissions differed by disease category, or was similar for all categories as would be expected if healthy vaccinee bias were the main explanation. Corroborating the findings from low-income countries, the effect of MMR was smallest for gastrointestinal infection admissions (HR=0.93 (0.87-1.00)), being stronger for lower respiratory infections (HR=0.80 (0.76-0.84); p=0.001 for same effect as for gastrointestinal infections) and for upper respiratory infections (HR=0.86 (0.82-0.89); p=0.058).
In the Dutch study, MMR+MenC was likewise associated with the smallest reduction in gastrointestinal infection admissions (0.69 (0.61-0.78)) whereas the reductions were significantly stronger for lower respiratory infections (HR=0.57 (0.49-0.65); p<0.046) and for upper respiratory infections (HR=0.36 (0.34-0.37); p<0.001). In contrast, DTP4 was associated with the same reductions in gastrointestinal (HR=0.70 (0.61-0.79)), lower respiratory (HR=0.79 (0.67-0.82)) and upper respiratory infection admissions (HR=0.72 (0.62-0.82)) (p=0.299 for same effect; estimates for >1 day admissions, as no estimates for all admissions were reported)). Thus, relative to the reduction for gastrointestinal infections, MMR+MenC was associated with stronger reductions in admissions for lower respiratory infections (HR=0.80 (0.65-0.98)) and upper respiratory infections (HR=0.77 (0.64-0.93)), and the effect on lower respiratory infections and upper respiratory infections differed significantly between MMR+MenC vaccination and DTP4 vaccination.
Thus, there are two observations from this triangulation analysis, which cannot be explained by healthy vaccinee bias: First, the pattern of stronger reductions for respiratory infections than for gastrointestinal infections after MMR(+/-MenC) vaccinations seen in both the Netherlands and Denmark. Second, the effect of MMR(+MenC) and DTP3/4 on respiratory infections differed significantly in both the Netherlands and Denmark, with beneficial effect estimates of MMR(+MenC) but the opposite tendency for DTP4. Thus, in spite of severe healthy vaccinee bias, there was still important information supporting beneficial NSEs of MMR+MenC on respiratory infections in the Dutch study. Reassuringly, a beneficial effect of MMR on respiratory infections has also recently been reported by studies from USA(5) and Italy(6).
There is increasing epidemiological and immunological evidence that NSEs of vaccines are important(2-7). The number of RCTs testing NSEs is going to be limited for ethical and financial reasons. Hence, we rely largely on observational studies to assess NSEs. As shown here, using triangulation methods it is possible to distinguish NSEs from bias, even in settings with severe healthy vaccinee bias. Hence, we hope the Dutch call for caution is not interpreted as a call for caution against observational studies of the NSEs of vaccines.
Christine S Benn, Signe Sørup, Peter Aaby
Research Centre for Vitamins and Vaccines (CVIVA), Bandim Health Project, Statens Serum Institut, Artillerivej 5, 2300 Copenhagen S; OPEN, Department of Clinical Research, and Danish Institute for Advanced Study, University of Southern Denmark/Odense University Hospital, Denmark
References
1. Tielemans SMAJ, de Melker HE, Hahné SJM, Boef AGC, van der Klis FRM, Sanders EAM, van der Sande MAB, Knol MJ. Non-specific effects of measles, mumps, and rubella (MMR) vaccination in high income setting: population based cohort study in the Netherlands. BMJ 2017;358:j3862
2. Sørup S, Benn CS, Poulsen A, Krause TG, Aaby P, Ravn H. Live vaccine against measles, mumps, and rubella and the risk of hospital admissions for nontargeted infections. JAMA 2014;311:826-35
3. Higgins JPT, Soares-Weiser K, López-López JA, Kakourou A, Chaplin K, Christensen H, Martin NK, Sterne JAC, Reingold AL. Association of BCG, DTP, and measles containing vaccines with childhood mortality: systematic review. BMJ 2016;355:i5170
4. Martins CL, Benn CS, Andersen A, Balé C, Schaltz-Buchholzer F, Do VA, Rodrigues A, Aaby P, Ravn H, Whittle H, Garly ML. A randomized trial of a standard dose of Edmonston-Zagreb measles vaccine given at 4.5 months of age: effect on total hospital admissions. J Inf Dis 2014;209;1731-8
5. Bardenheier BH, McNeil MM, Wodi AP, McNicholl J, DeStefano F. Risk of nontargeted infectious disease hospitalizations among U.S. children following inactivated and live vaccines, 2005-2014. Clin Inf Dis 2017;epub
6. La Torre G, Saulle R, Unim B, Meggiolaro A, Barbato A, Mannocci A, Spadea A. The effectiveness of measles-mumps-rubella (MMR) vaccination in the prevention of pediatric hospitalizations for targeted and untargeted infections: a retrospective cohort study. Human Vaccines & Immunotherapeutics 2017; 13: 1879-83
7. Benn CS, Netea MG, Selin LK, Aaby P. A small jab - a big effect: nonspecific immunomodulation by vaccines. Trends Immunol. 2013 Sep;34(9):431-9.
Competing interests: No competing interests
Changes in sequence of vaccinations and changes in hospital admission patterns?
We appreciate that Dr. Knol and colleagues clarified that the hazard ratio (HR) of hospital admissions of 0.85 (95% CI: 0.70-1.03) for DTaP-IPV-Hib-4+PCV after MMR+MenC was for all admissions, whereas the estimate was 0.93 (95% CI: 0.67-1.31) for admissions with a duration of >1 day. However, we are sorry that they declined to report the estimate by the four disease admission categories used in the study, i.e. upper respiratory infections (URI), lower respiratory infections (LRI), gastrointestinal infections (GI) and “other infections”.
In the Dutch study, the children were called for vaccination and less than 1% did not received the MMR+MenC vaccinations scheduled for 14 months of age in the Netherlands. Hence, the reference groups with delayed routine vaccinations tended to be small frail subgroups. Accordingly, the main analysis of the Dutch cohort indicated strong healthy vaccinee bias in the transitions from DTaP-IPV-Hib-3+PCV to DTaP-IPV-Hib-4+PCV and from DTaP-IPV-Hib-4+PCV to MMR+MenC; MMR+MenC after DTaP-IPV-Hib-4+PCV was associated with hazards ratios (HRs) of 0.62 (0.57-0.67) for admissions>1 day and 0.40 (0.38-0.41) for all admissions.
Using these estimates as a way to adjust for the potential healthy vaccinee biases in the Dutch vaccination system, the differential effect of having DTaP-IPV-Hib-4+PCV after MMR+MenC rather than MMR+MenC after DTaP-IPV-Hib-4+PCV would be a 2.13 (1.75-2.59) fold increase in the relative risk for all admissions (0.85/0.40); for admission>1 day, the differential effect would be a 1.50 (1.06-2.12) fold increase in the relative risk (0.93/0.62).
In our Danish study, the children were not called for vaccination, the final coverage for MMR at 24 months of age was only 87%, and thus there was much less healthy vaccinee bias. In Denmark, the transition from DTaP-IPV-Hib-3 to MMR was associated with an HR of 0.86 (0.84-0.88) for all admissions, whereas the transition from MMR to DTaP-IPV-Hib-3 was associated with a HR of 1.62 (1.28-2.05). Hence, compared with the transition to routine MMR from DTaP-IPV-Hib-3, receiving DTaP-IPV-Hib-3 after MMR increased the relative admission risk 1.86-fold (1.62/0.86) - very similar to the effect in the Dutch study.
In the Danish study, the increase in admissions for children having received DTaP-IPV-Hib-3 after MMR was particularly strong for URI (HR=1.89 (1.31-2.74)) and LRI (HR=1.64 (1.06-2.54)), whereas there were only insignificant increases in GI and “other infections”. This is interesting because it is the opposite pattern of what happened when MMR was given after DTaP-IPV-Hib-3; here the reduction in admissions was particularly strong for LRI and URI.
In the Dutch study, the reductions in admission for MMR+MenC after DTaP-IPV-Hib-4+PCV were also particularly strong for LRI and URI for both all admissions and admissions>1 day. In contrast, when DTaP-IPV-Hib-4+PCV followed DTaP-IPV-Hib-3+PCV there was an overall reduction in admissions (HR=0.69 (0.63-0.76)) presumably reflecting a healthy vaccinee bias, but there was less reduction for URI (HR=0.72) and LRI (HR=0.79) than for GI (HR=0.70) and “other infections” (HR=0.59). Compared with the effect of MMR+MenC, the DTaP-IPV-Hib-4+PCV was associated with a relative increase in URI and LRI (rapid responses). In other words, the Dutch data show the same differential effect on morbidity patterns as the Danish data; relative to the effects of MMR, the non-live DTaP-IPV-Hib is associated with a relative increase in URI and LRI. Hence, it would also be very interesting to examine in the Dutch data what happens when DTaP-IPV-Hib-4+PCV is given after MMR.
This is important. In low-income countries, if doses of DTP are missing it is WHO policy to provide the missing doses even if it is after measles vaccine (MV). However, all studies which have examined the sequence have found that receiving DTP after MV is associated with considerable increased mortality compared with having MV as the most recent vaccine. In the WHO meta-analysis of non-specific effects this sequence of vaccinations was associated with a HR of 2.66 (1.04-6.81) (Higgins et al, BMJ 2016). If DTP-containing vaccines compared with MV-containing vaccines enhance the susceptibility to respiratory infections this might be one of the explanation for this increased mortality.
Dr. Knol and colleagues suggest that non-specific effects and healthy vaccinee bias cannot be disentangled in observational studies, and recommend RCTs to obtain definite answers. The fact that DTaP-IPV-Hib-4+PCV and MMR+MenC are followed by different morbidity patterns is in itself an indication that these vaccines have differential non-specific effects (not differential healthy vaccinee bias), though we may not find the “true” magnitude of this effect in an observational study. The likelihood that anyone will get funding and permission to conduct RCTs of different sequences of vaccinations is limited. Hence, it is important that we collect as much information as possible from observational studies to justify at least some RCTs. We therefore hope that the Dutch groups will provide their estimates for DTaP-IPV-Hib-4+PCV after MMR+MenC by disease group for all admissions and admissions>1 day.
Competing interests: No competing interests