SEARCH

SEARCH BY CITATION

Keywords:

  • non-specific effects;
  • vaccines;
  • measles;
  • diphtheria-tetanus-pertussis;
  • mortality
  • effets non spécifiques;
  • vaccins;
  • rougeole;
  • diphtérie-tétanos-coqueluche;
  • mortalité
  • efectos no específicos;
  • vacunas;
  • sarampión;
  • difteria-tétano-pertusis;
  • mortalidad

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and subjects
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Appendix

Objectives  Studies from low-income countries have suggested that routine vaccinations may have non-specific effects on child mortality; measles vaccine (MV) is associated with lower mortality and diphtheria-tetanus-pertussis (DTP) with relatively higher mortality. We used data from Navrongo, Ghana, to examine the impact of vaccinations on child mortality.

Methods  Vaccination status was assessed at the initiation of a trial of vitamin A supplementation and after 12 and 24 months of follow-up. Within the placebo group, we compared the mortality over the first 4 months and the full 2 years of follow-up for different vaccination status groups with different likelihoods of additional vaccinations during follow-up. The frequency of additional vaccinations was assessed among children whose vaccination card was seen at 12 and 24 months of follow-up.

Results  Among children with a vaccination card, more than 75% received missing DTP or MV during the first 12 months of follow-up, whereas only 25% received these vaccines among children with no vaccination card at enrolment. Children without a card at enrolment had a significant threefold higher mortality over the 2-year follow-up period than those fully vaccinated. The small group of children with DTP3-4 but no MV at enrolment had lower mortality than children without a card and had the same mortality as fully vaccinated children. In contrast, children with 1–2 DTP doses but no MV had a higher mortality during the first 4 months than children without a card [MRR = 1.65 (0.95, 2.87)]; compared with the fully vaccinated children, they had significantly higher mortality after 4 months [MRR = 2.38 (1.07, 5.30)] and after 2 years [MRR = 2.41 (1.41, 4.15)]. Children with 0–2 DTP doses at enrolment had higher mortality after 4 months (MRR = 1.67 (0.82, 3.43) and after 2 years [MRR = 1.85 (1.16, 2.95)] than children who had all three doses of DTP at enrolment.

Conclusions  As hypothesised, DTP vaccination was associated with higher child mortality than measles vaccination. To optimise vaccination policies, routine vaccinations need to be evaluated in randomised trials measuring the impact on survival.

Objectifs:  Des études provenant des pays à faibles revenus suggèrent que les vaccinations de routine pourraient avoir des effets non spécifiques sur la mortalité infantile; le vaccin contre la rougeole (VR) a été associée à une mortalité plus faible et le vaccin diphtérie-tétanos-coqueluche (DTC) à une mortalité relativement élevée. Nous avons utilisé les données de Navrongo, au Ghana, pour examiner l’impact des vaccinations sur la mortalité infantile.

Méthodes:  Le statut vaccinal a étéévalué lors de l’initiation d’un essai clinique de supplémentation en vitamine A et après 12 et 24 mois de suivi. Dans le groupe placebo, nous avons comparé la mortalité au cours des 4 premiers mois et au cours des 2 années complètes de suivi pour les différents groupes de statut vaccinal avec des chances différentes pour des vaccinations additionnelles au cours du suivi. La fréquence des vaccinations additionnelles a étéévaluée chez les enfants dont la carte de vaccination a été contrôlée à 12 et 24 mois de suivi.

Résultats:  Parmi les enfants ayant un carnet de vaccination, plus de 75% ont reçu une vaccination DTC ou VR manquée, au cours des 12 premiers mois de suivi, alors que seulement 25% ont reçu de tels vaccins chez les enfants sans carnet de vaccination lors de l’inclusion. Les enfants sans carte lors de l’inclusion avaient une mortalité significativement 3 fois plus élevée au cours de la période de 2 ans de suivi que ceux qui étaient entièrement vaccinés. Le petit groupe d’enfants ayant reçu le vaccin DTP3-4, mais pas le VR lors de l’inclusion avait une mortalité plus faible que les enfants sans carte et avait le même taux de mortalité que les enfants complètement vaccinés. Par contre, les enfants ayant reçu 1à 2 doses de DTC mais pas de VR avaient une mortalité plus élevée au cours des 4 premiers mois que les enfants sans carte (MRR = 1,65 (0,95–2,87)); comparés aux enfants complètement vaccinés ils avaient une mortalité significativement plus élevée après 4 mois (MRR = 2,38 (1,07–5,30)) et après 2 ans (MRR = 2,41 (1,41–4,15)). Les enfants ayant reçu 0 à 2 doses de DTC à l’inclusion avaient une mortalité plus élevée après 4 mois (MRR = 1,67 (0,82–3,43) et après 2 ans (MRR = 1,85 (1,16–2,95)) que les enfants qui avaient reçu toutes les trois doses de DTC lors de l’inclusion.

Conclusions:  Comme prédit, le vaccin DTC a été associéà une mortalité plus élevée que le vaccin contre la rougeole. Afin d’optimiser les politiques de vaccination, les vaccinations de routine devraient être évaluées dans des essais randomisés pour mesurer l’impact sur la survie.

Objetivos:  Estudios de países con ingresos bajos sugieren que las vacunaciones rutinarias podrían tener efectos no específicos sobre la mortalidad infantil; la vacuna del sarampión (VS) está asociada con una menor mortalidad y la difteria-tetanus-pertussis (DTP) con una mortalidad relativamente mayor. Hemos utilizado datos de Navrongo, Ghana, para examinar el impacto de las vacunas sobre la mortalidad infantil.

Métodos:  El estatus vacunal se evaluó al comienzo del ensayo de suplementación con vitamina A y después de 12 y 24 meses de seguimiento. Dentro del grupo placebo, hemos comparado la mortalidad durante los 4 primeros meses y los 2 años enteros de seguimiento, de grupos con diferente estatus de vacunación y diferente probabilidad de recibir vacunas adicionales durante el seguimiento. La frecuencia del vacunas adicionales se evaluó entre aquellos niños cuyo carnet vacunal estuvo disponible en los meses 12 y 24 del seguimiento.

Resultados:  Entre los niños con carnet vacunal, más de un 75% recibieron, durante los 12 primeros meses del seguimiento, la(s) dosis de DTP o VS que les faltaban, comparado con solo un 25% de los niños que no tenían el carnet vacunal al ingresar en el estudio. Los niños sin carnet vacunal en el momento de ingresar en el estudio tenían una mortalidad significativa, 3 veces mayor que aquellos que estaban completamente vacunados, durante el periodo de 2 años de seguimiento. El pequeño grupo de niños con DTP3-4 pero sin VS al momento de ingresar en el estudio tenían una menor mortalidad que los niños sin carnet vacunal, y la misma mortalidad que los niños con todas la vacunas. En contraste, los niños con las dosis 1-2 DTP pero sin VS tenían una mayor mortalidad durante los 4 primeros meses que los niños sin un carnet vacunal (MRR=1.65 (0.95, 2.87)); comparados con los niños con vacunación completa, tenían una mortalidad significativamente más alta después de 4 meses (MRR=2.38 (1.07,5.30)) y después de 2 años (MRR=2.41(1.41,4.15)). Los niños con 0-2 dosis de DTP al comienzo del estudio tenían una mayor mortalidad después de 4 meses (MRR=1.67(0.82,3.43) y después de 2 años (MRR=1.85(1.16,2.95)) que los niños que tenían las tres dosis de DTP al comienzo del estudio.

Conclusiones:  La vacunación con DTP estaba asociada a una mayor mortalidad infantil que la vacunación para el sarampión. Con el fin de optimizar las política de vacunación, deberían evaluarse las vacunas rutinarias en ensayos aleatorizados que midan el impacto sobre la supervivencia.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and subjects
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Appendix

Routine childhood vaccinations may have non-specific effects (NSE) on morbidity and mortality in low-income countries with a high burden of infectious diseases (Aaby et al. 1995, 2003b; Kristensen et al. 2000). The best known example is the studies of high-titre measles vaccine (HTMV), which was associated with twofold higher mortality for girls than the standard measles vaccine (MV). HTMV was fully protective against measles infection, and the effect was therefore a non-specific effect (Aaby et al. 2003b). Such NSE are presumably owing to modulation of the immune system. The experience with HTMV suggested that NSE could have major effects on child survival. This finding made us investigate whether other vaccines could have NSE.

Using a landmark approach to survival analysis (Aaby et al. 2007; Jensen et al. 2007), studies showed that the common vaccinations used in infancy had differential NSEs; BCG (Roth et al. 2006a,b) and MV were associated with lower subsequent mortality (Aaby et al. 1995) and diphtheria-tetanus-pertussis (DTP) with relatively higher subsequent mortality (Kristensen et al. 2000; Aaby et al. 2004a,b; Veirum et al. 2005). These results have been controversial because they question basic assumptions of the current vaccination programme and suggest that changes in the programme might be needed (Fine & Smith 2007).

More data sets from other countries are therefore needed to assess the magnitude of NSE and whether different routine vaccinations have different NSE. Surprisingly, few longitudinal studies in low-income countries have collected routine information on vaccinations. The Ghana vitamin A supplementation trial (Ghana VAST)’s Survival Study in Kassena-Nankana District, Ghana, in 1989–1991 assessed vaccination status at the initiation of the trial. Previously, we have demonstrated that the impact of vitamin A supplementation (VAS) may be modified by interaction with vaccines (Benn et al. 2009). In the present analysis, we have focused on the impact of vaccines on child mortality among children who received placebo to study the impact of vaccines in the absence of VAS.

Methods and subjects

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and subjects
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Appendix

Ghana VAST Survival Study

VAST was conducted between 1989 and 1991 (Ghana VAST Study Team 1993; Binka et al. 1995; Ross et al. 1995; Benn et al. 2009). The details of this cluster randomised trial of vitamin A supplementation (VAS) have been described in the previous publications and the Appendix. Dosing was every 4th month. Over the 2 years, 21 906 children were enrolled in the trial by receiving at least one dose of either VAS or placebo, 13 462 being enrolled in Round 1. VAS was associated with a 19% (95%CI 2–32%) reduction in mortality.

Information on vaccinations

Vaccination information was collected systematically for the children enrolled in Round 1. One year and 2 years later, in Rounds 4 and 7, vaccination information was collected again. Owing to an archiving error, dates of measles vaccination from Round 7 were not saved in the stored data files. The present analysis has been restricted to children enrolled in Round 1 for whom there was follow-up for the whole 2-year period; we only included children who had received placebo since this group would not be subject to vaccine/vitamin A interactions.

Data on vaccinations were collected de novo each time without taking previous information into consideration and only for children being alive at the visit. Thus, the vaccination status of children who died between vaccination visits was not updated. Comparing information from rounds 1 and 4 for the same children, there were minor inconsistencies (Appendix).

Follow-up and verbal autopsies

All children in the study were visited every 4 months. Deaths were identified through these visits and independently by key informants based in the community. With the intensive follow-up and the need to report, the status of all children in the study it is unlikely that deaths were missed. The verbal autopsy procedure for all deaths makes it unlikely that surviving children were erroneously classified as dead. The follow-up procedures were independent of the vaccination status of the child.

When a child had died an attempt was made to complete a verbal autopsy questionnaire with the mother/guardian to ascertain the circumstances of the death and the symptoms associated with the final illness. No information on vaccines was collected. Questionnaires were read and classified by two physicians (DAR, Fred Binka). If any of the physicians classified a death as because of measles infection, the death has been classified as measles in the present analysis.

Background factors and nutritional status

Information on a number of socio-economic (SES) and demographical background factors, feeding mode, vaccination history, recent and past morbidity and anthropometry were collected at the beginning of the study. Weight and arm circumference (MUAC) was assessed in the same rounds as vaccination information. We calculated weight-for-age (WAZ) and MUAC z-scores using the WHO standards (http://www.who.int/childgrowth/software/en). We used a z-score cut-off of < −2 in analyses controlling for malnutrition.

Analysis of vaccinations

Information from Rounds 1, 4 and 7 was used to calculate the proportion of children who received vaccinations in the years following enrolment. This proportion was estimated among children whose health card had been seen at enrolment and again 1 year or 2 years later, that is, among surviving children. The incidence of vaccination was very high in the first months after enrolment, because of a major intensification of the vaccination programme before and immediately after 1990, the year when the Expanded Programme on Immunisation had targeted achieving 80% coverage of vaccination globally (Kim-Farley 1992). Therefore, a certain vaccination status in Round 1 or Round 4 not only reflected which vaccines the child had already received, but also which vaccines they were likely to receive in the coming months. Hence, any vaccination effects on mortality are likely to be a product of both vaccination status at enrolment and the vaccines likely to be received during follow-up. We have particularly focused on the impact on survival in the first 4 months when the information on immunisation status would be most valid. However, we have also reported the mortality effect for the full 2 years duration of the trial. For the present analyses, children who had no health card at Round 1 have been assumed to be ‘unvaccinated’.

We have emphasised two aspects of vaccination status: first, children who had not yet received MV were likely to receive MV during follow-up; second, children who had not received all doses of DTP were likely to receive additional doses of DTP during follow-up. We have therefore distinguished between children who had received 0–2 doses (DTP0-2) or 3 or more doses of DTP (DTP3-4) at enrolment. As indicated in Table 2, vaccination status at enrolment was categorised into 6 groups.

Statistical analyses

Comparisons of background factors for children with and without a health card were carried out using logistic regression, linear regression or nonparametric tests. For the assessment of the risk of dying within 4 months or 2 years, the period at risk was defined as the time period between the enrolment date and either the date the child died or was censored or 4 months/2 years later, whatever came first. The survival data were analysed in Cox-proportional hazards models with age as the underlying time and reported as mortality rate ratios (MMRs) with 95% confidence interval (CI). We also analysed the data using time since visit as the timescale in a Cox model and controlled for age as a categorised variable (6–11, 12–23 and 24–35 months). The results are essentially the same as in the Cox model using age as the timescale (see Appendix). Cox-proportional hazard assumptions were tested and were not violated. Background factors associated with a < 0.10 were included in the final model using backward selection. The study area was divided into four geographical Zones: North, South, East and West. As the WAZ score was better for children who did not have a vaccination card – possibly due to less accurate age assessment (see Appendix) – we also controlled for WAZ in all analyses. Data were analysed in Stata 11.

To minimise the effect of survival bias, we used a landmark approach with vaccinations status at enrolment being a fixed time variable during follow-up (Jensen et al. 2007). Survival bias leads to differential misclassification of vaccine status because children who survive have better information than dead children because health cards for dead children were often not available for vaccination information to be updated. This makes vaccine information for the dead children incomplete making vaccination to be automatically associated with a strong beneficial effect.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and subjects
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Appendix

Study population

A total of 6882 children were enrolled in the placebo group in Round 1; 49% had a health card seen, 13% were said to have a card but it was not seen, 37% did not have a card, and 0.4% had no information recorded on their health card status. The present analysis is restricted to children whose card was seen (3397) or who had no card (2577). Within the first 4 months 91 (1.5%) of these children died and over the 2 years of the trial, 274 (4.6%) children died; 51/274 (19%) were because of measles and the two physicians agreed on 47 of 51 diagnoses of measles.

Vaccination coverage

Fewer than 50% of the children had received BCG, DTP and MV (Table 1). For those with a health card seen, however, the coverage continued to increase until around 24 months of age. Most children had received their routine vaccinations out-of-sequence; 76% had received BCG and DTP simultaneously, and 86% (1771/2057) of those who had received both MV and DTP had received MV and DTP simultaneously or at least one DTP after MV.

Table 1. Vaccination coverage at enrolment into the trial according to age in the placebo group
VaccineCoverage of vaccinesCoverage by age among children with a health card n (%)
All childrenChildren with card
All5974 n (%)3397 n (%)5–11 months N = 358 n (%)12–17 months N = 484 n (%)18–23 months N = 409 n (%)24–35 months N = 738 n (%)36+ months N = 1408 n (%)
  1. *Percentage only among children with a vaccine.

BCG2639 (44)2639 (78)305 (85)409 (85)353 (86)589 (80)983 (70)
BCG with DTP* 2003 (76)*258 (85)341 (83)291 (82)466 (79)647 (66)
Any DTP2795 (47)2795 (82)308 (86)436 (90)368 (90)636 (86)1047 (74)
DTP3/DTP41503 (25)1503 (44)87 (24)183 (38)191 (47)378 (51)664 (47)
Any OPV2781 (47)2781 (82)312 (87)424 (88)369 (90)629 (85)1047 (74)
OPV3-51426 (24)1426 (42)80 (22)165 (34)180 (44)366 (50)635 (45)
MV2208 (37)2208 (65)86 (24)266 (55)307 (75)556 (75)993 (71)
MV with DTP* 1352 (66)*64 (76)179 (70)206 (69)374 (71)529 (60)
DTP after MV* 1096 (53)*8 (10)73 (28)132 (44)294 (55)589 (66)

Among children aged 3 years or older at enrolment, subsequent vaccination intensity was limited. In the following analysis, we have therefore focused on children under 3 years of age in the placebo group (N = 3082; Figure 1).

image

Figure 1.  Study population of children aged 6–35 months at enrolment into the placebo group.

Download figure to PowerPoint

Vaccinations during follow-up

The vaccination intensity in the first 4 months after enrolment was high for children with a health card (Table 2). During the first 4 months, 50% (408/817) received additional doses of DTP if they had a vaccination card but had not received DTP3 at enrolment (Table 2, Group 2, 3 and 5); 76% (621/817) received DTP within 12 months. Among children who had not had MV at enrolment, 50% (276/549) received this vaccine during the next 4 months (Table 2, Group 2, 3 and 4); 79% (431/549) received MV within 12 months. In contrast, children who had received DTP3-4 before enrolment (Table 2, Group 4 and 6) were unlikely to receive additional doses of DTP during follow-up [6% (39/663)]. Among children not having a health card at enrolment, only 25% received at least one dose of DTP or MV during the first 12 month of follow-up (Table 2, Group 1).

Table 2. Proportion of children receiving vaccination during follow-up, according to vaccination status at enrolment, among children <36 months of age at enrolment into the placebo group
 Age (months) at entry n (%)% Received DTP (N)% Received MV (N)
Vaccination status at enrolment6–1112–2324–35First 4 monthsFirst year2 yearsFirst 4 monthsFirst year
  1. Group 1 consisted of those with no health card and was considered ‘unvaccinated’. Groups 2–6 all had health cards at Round 1. Group 2 had not received DTP or MV (some of these children had, however, received BCG); Group 3 had received DTP1-2 but not MV; Group 4 had received DTP3-4, but not MV; Group 5 had received DTP1-2 and MV; and Group 6 had received DTP3-4 and MV (fully vaccinated).

  2. *DTP vaccinations after MV.

1No health card265 (24.3)18 (152/863)25 (214/863)24 (169/714)14 (124/863)25 (218/863)
409 (37.4)
419 (38.3)
2Health card (no DTP, no MV)48 (24.5)59 (73/123)78 (96/123)81 (87/108)45 (55/123)77 (95/123)
72 (36.7)
76 (38.8)
3Health card + DTP1/2 (no MV)182 (38.8)57 (191/335)81 (271/335)83 (250/301)47 (158/335)78 (262/335)
201 (42.9)
86 (18.3)
4Health card + DTP3/4 (no MV)42 (38.5)10 (9/91)11 (10/91)13 (10/80)69 (63/91)81 (74/91)
47 (43.1)
20 (18.4)
5Health card + MV and DTP0-241 (8.5)40 (144/359)*71 (254/359)*74 (233/315)*NANA
246 (50.7)
198 (40.8)
6Health card + MV and DTP3-445 (6.2)5 (30/572)*6 (33/572)*5 (28/514)*NANA
327 (44.8)
358 (49.0)

Background factors

The children with and without a health card were different in several respects (Table 3). Controlling for age, the children with a vaccination card were more likely to still be breastfed. The children with a health card had lower weight-for-age z-scores (WAZ) than children with no card. The difference in WAZ could be due to current age of the children without a health card having been underestimated. Also, children with a card were more likely to have been hospitalised, presumably indicating that communities with high coverage had better contact with health services. Fifteen per cent of the children without a health card had a BCG scar. These children may have lost the card or never been issued with one.

Table 3. Background characteristics of children with and without a health card, among children <36 months of age at enrolment into the placebo group
 Health cardNo health cardOdds ratio/mean difference by background characteristics (95% CI)
N = 1989 (65%) N = 1093 (35%) N = 3082
  1. IQR, Inter Quartile Range.

  2. *Kruskal–Wallis test.

Sex (Males/N)49.5% (984/1989)50.9% (557/1093)0.94 (0.81, 1.09)
Median age (months)20.0 [IQR 13.6; 27.5]19.6 [IQR 12.3; 28.0]  P = 0.163*
Ever breastfed100% (1989/1989)99.9% (1092/1093)P = 0.164*
Still breastfeeding81.0% (1612/1989)75.4% (824/1093)1.40 (1.16, 1.67)
Mean arm circumference MUAC z-scores−1.47 (−1.52; −1.41)−1.40 (−1.47; −1.33)−0.07 (−0.16, −0.02)
Mean weight-for-age z-score−1.94 (−2.00; −1.88)−1.67 (−1.76; −1.58)−0.27 (−0.37, −0.16)
BCG scar80.9% (1609/1988)15.0% (164/1093)24.0 (19.6, 29.5)
Ever had measles before enrolment2.0% (39/1989)1.9% (21/1091)1.02 (0.58, 1.83)
Previously admitted to hospital4.6% (92/1989)2.5% (27/1093)1.91 (1.23, 3.08)
Radio in compound31.3% (622/1985)18.6% (202/1089)2.0 (1.67, 2.41)

Mortality of different vaccination groups

In the first 4 months, the MRR for children with a health card compared with children without a card and adjusted for zone, WAZ and ownership of a radio was 0.97 (95%CI 0.61, 1.56); 1.10 (95%CI 0.56, 2.16) for vaccinated girls and 0.86 (95%CI 0.44, 1.69) for vaccinated boys. Over the 2 years of follow-up, the children with a card had significantly lower mortality than children without a card [MRR = 0.61 (0.46, 0.82)].

Children who had received MV by enrolment had slightly better WAZ than children who had received at least one dose of DTP but not MV (DTP-vaccinated children), the RR of WAZ < −2 being 0.93 (95%CI 0.88, 1.00) for MV children. Controlled for WAZ, over the first 4 months, the children who were MV-vaccinated at enrolment had half the mortality of the children who were initially only DTP-vaccinated [MRR = 0.51 (95%CI 0.27, 0.97)] (Table 4). The reduction in mortality could not be explained by protection against measles deaths, as defined by the verbal autopsies, since the MRR was 0.46 (95%CI 0.23, 0.89) when measles deaths were excluded from the analysis (Table 4).

Table 4. Mortality rate ratio during the first 4 months of follow-up for children with MV at enrolment vs. children with DTP without MV at enrolment among children aged 6–35 months at enrolment into the placebo group
Vaccination status at enrolmentMR per 1000 years (Deaths/person-years) [children enrolled] {measles deaths}
BoysGirlsAll
  1. MRR, Mortality Rate Ratio.

  2. *Adjusted for age, zone, weight-for-age and ownership of radio.

MV at enrolment (Table 2, Group 5 and 6)33 (7/209) [593] {1}45 (10/220) [622] {1}40 (17/429) [1215] {2}
DTP (No MV at enrolment) (Table 2, Group 3 and 4)127 (13/102) [289]139 (14/101) [289]133 (27/203) [578]
MRR MV/DTP (No MV at enrolment)*0.50 (0.19, 1.31)0.52 (0.22, 1.23)0.51 (0.29, 0.97)
MRR* (measles deaths excluded)0.44 (0.16, 1.22)0.48 (0.20, 1.16)0.46 (0.23, 0.89)

As vaccination status group at enrolment influenced the likelihood of receiving additional vaccinations during follow-up (Table 2), we also assessed the relative mortality of these different vaccination status groups (Table 5). Fully vaccinated children (Group 6) had lower mortality than children without a card (Group 1), and this effect was significant for the 2-year follow-up period [MRR = 0.37 (95%CI 0.23, 0.62)]. The small group of children with DTP3-4 but no MV at enrolment (Group 4) had similar mortality as fully vaccinated children (Group 6), and roughly half the mortality rate compared with the children without a card both after 4 months and 2 years, though neither of these differences were statistically significant (Table 5). Hence, missing vaccines was not per se a risk factor for child mortality. The prevention of measles deaths did not appear to explain the lower mortality among the measles-vaccinated children and the children with DTP3-4 but no MV (Table 5).

Table 5. Mortality rate (MR) per 1000 person-years (pyrs) and mortality rate ratio (MRR) for vaccinated compared with unvaccinated children, among children <36 months by enrolment into the placebo group
Group: Vaccination status at enrolment N Median weight-for-age z-score at enrolmentMR per 1000 years (deaths/pyrs) {measles deaths}Within 4-month follow-upWithin 24-month follow-up:
Unadjusted MRR (95% CI) {MRR – without measles deaths)MRR* (95% CI) (Vaccinated/No heath card) {MRR – without measles deaths}MR per 1000 years (deaths/pyrs) {measles deaths}Unadjusted MRR (95% CI) {MRR – without measles deaths)MRR* (95% CI) (Vaccinated/No health card) {MRR – without measles deaths}
  1. *Adjusted for age, zone, weight-for-age and ownership of radio (N = 3082).

1. No card1093−1.7 [−2.6; −0.8] 80 (31/389) {2}1.0 (Reference)1.0 (Reference)51 (101/1998) {23}1.0 (Reference)1.0 (Reference)
2. Health card:  noDTP, noMV196−1.9 [−2.9; −1.0] 58 (4/69)0.72 (0.25, 2.05) {0.77 (0.27, 2.19)}0.69 (0.24, 1.97) {0.75 (0.26, 2.13)}30 (11/370) {2}0.59 (0.32, 1.10) {0.63 (0.31, 1.25)}0.55 (0.29, 1.02) {0.59 (0.29, 1.17)}
3. DTP1-2, no MV469−2.0 [−3.0; −1.2]152 (25/164)1.90 (1.12, 3.22) {2.03 (1.19, 3.47)}1.65 (0.95, 2.87) {1.74 (1.00, 3.05)}54 (46/854) {6}1.07 (0.75, 1.51) {1.19 (0.82, 1.75)}0.90 (0.63, 1.30) {1.03 (0.69, 1.52)}
4. DTP3-4, no MV109−1.8 [−3.0; −1.2] 52 (2/38)0.66 (0.16, 2.74) {0.70 (0.17, 2.94)}0.57 (0.13, 2.42) {0.61 (0.14, 2.61)}24 (5/207) {1}0.48 (0.20, 1.19) {0.50 (0.18, 1.36)}0.44 (0.18, 1.08) {0.46 (0.17, 1.28)}
5. MV, DTP0-2485−1.9 [−2.8; −1.1] 47 (8/171){2}0.58 (0.27, 1.27) {0.47 (0.19, 1.23)}0.80 (0.36, 1.78) {0.66 (0.27, 1.61)}26 (24/923) {3}0.52 (0.33, 0.81) {0.58 (0.36, 0.94)}0.62 (0.39, 0.97) {0.75 (0.46, 1.23)}
6. MV, DTP3-4730−1.8 [−2.6; −1.0] 35 (9/257)0.44 (0.21, 0.92) {0.47 (0.22, 0.98)}0.70 (0.32, 1.53) {0.76 (0.34, 1.69)}14 (20/1408)0.28 (0.18, 0.46) {0.36 (0.22, 0.60)}0.37 (0.23, 0.62) {0.51 (0.30, 0.85)}
Total3082−1.8 [−2.7; −1.0] 73 (79/1090){4}  36 (207/5760){35}  

In contrast, children with 1–2 doses of DTP but no MV at enrolment (Table 5, Group 3) had a higher overall mortality than children without a card (Group 1) during the first 4 months of follow-up [MRR = 1.65 (95%CI 0.95, 2.87)], with mortality being significantly higher for deaths that were not attributed to measles [MRR = 1.74 (95%CI 1.00, 3.05)]. The DTP1-2-vaccinated children (Group 3) also had a twofold higher mortality rate than the fully vaccinated children (Group 6) in the first 4 months [MRR = 2.38 (95%CI 1.07, 5.30)] and over the 2 years [MRR = 2.41 (95%CI 1.41, 4.15)].

Controlling for MV status (Table 6), we compared children with missing doses of DTP (Table 5, Groups 2, 3 and 5) with children who had already received DTP3-4 at enrolment (Table 5, Groups 4 and 6). Those missing doses of DTP tended to have higher mortality during the first 4 months [MRR = 1.67 (95%CI 0.82, 3.43)] and had significantly higher mortality over the full duration of the study [(MRR = 1.85 (95%CI 1.16, 2.95)] than those with DTP3-4 at enrolment (Table 6).

Table 6. Mortality during follow-up according to number of doses of DTP received before enrolment, among children aged 6-35 months who had a health card at enrolment into the placebo group
Number of doses of DTP at enrolmentNo MV at enrolment (Table 2, Group 2, 3 and 4) MR per 1000 (Deaths/pyrs) [children enrolled] {measles deaths}Had received MV at enrolment (Table 2, Group 5 and 6) MR per 1000 (Deaths/pyrs) [children enrolled] {measles deaths}
  1. *Adjusted for age, zone, weight-for-age and ownership of radio and MV status at enrolment.

4 months of follow-up
 DTP0-2124 (29/234) [665] {0}47 (8/171) [485] {2}
 DTP3-453 (2/38) [109]35 (9/257) [730]
 MRR (DTP0-2/DTP3-4)* [without measles deaths]2.29 (0.53, 9.78) [2.29 (0.53, 9.78)]1.08 (0.38, 3.06) [0.88 (0.30, 2.59)]
 MRR (DTP0-2/DTP3-4)* [without measles deaths]1.67 (0.82, 3.43) [1.60 (0.78, 3.28)] 
24 months of follow-up
 DTP0-247 (57/1224) [665] {8}26 (24/923) [485] {3}
 DTP3-424 (5/207) [109] {1}14 (20/1408) [730]
 MRR (DTP0-2/DTP3-4)* [without measles deaths]1.89 (0.75, 4.77) [1.93 (0.69, 5.41]1.70 (0.91, 3.17) [1.53 (0.81, 2.90]
 MRR (DTP0-2/DTP3-4)* [without measles deaths]1.85 (1.16, 2.95) [1.68 (1.03, 2.73)] 

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and subjects
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Appendix

The data from Navrongo provided several indications that vaccines may have important NSE on child survival. The children who had received MV at enrolment had twofold lower mortality than children who had received only DTP and this difference was not explained by the prevention of measles deaths. Furthermore, adjusted for MV vaccination status at enrolment, those with incomplete doses of DTP had higher mortality than the children who had already received DTP3 before enrolment. These trends are consistent with observations made in several other studies from Africa (Aaby et al. 1995, 2006a,b,c; Kristensen et al. 2000).

Strengths and weaknesses

This was an observational study of the effects of vaccination. Misclassification of vaccination status did occur. A small proportion of vaccinated children were probably wrongly classified as unvaccinated. Information on vaccination dates may also have been missed or noted wrongly for a few of the children. Assuming these misclassifications to be random, such errors should tend to make the estimated differences more conservative.

The Navrongo data set is special in several ways. First, the study only included children older than 6 months and there was very little follow-up time and only three deaths before 9 months of age when BCG and DTP will usually have their strongest NSE. Hence, the DTP vaccines administered in the present study were likely to have been given out-of-sequence, that is, DTP co-administered with MV or DTP given after MV. The study is therefore less relevant for assessing the impact of the primary series of DTP vaccinations administered before 9 months of age. Second, in Navrongo a large part of the population remained unvaccinated at that time, and therefore, it was possible to compare vaccinated with unvaccinated children over a wide age range; the unvaccinated children were not merely a small particularly frail subgroup, which had been too weak to get vaccinated. Third, in other studies, we have analysed the impact of a vaccine, while it was presumed to be the latest vaccine received (Aaby et al. 2004b,c, 2006a,b,c; Veirum et al. 2005). This was not possible in this study because the time interval for collection of vaccination data was too long. Hence, vaccinations were evaluated as risk factors for mortality taking into consideration the initial vaccination status and whether additional vaccines may have been given during follow-up.

It could be speculated that the increased mortality associated with DTP merely represented poor social conditions, disorganised parenting, differential ascertainment of vaccinations or differential access to the EPI programme. However, the DTP-vaccinated children would have had better families or more compliant parents than the unvaccinated children. Nonetheless, the DTP-vaccinated children had higher mortality than the unvaccinated children. The late 1980s when these data were collected was the introduction period for the EPI programme, and many vaccines were given in campaigns and out-of-sequence. The process of who got vaccinated when are likely to have been much more random depending on when there was a drive to organise vaccinations; for example, the incidence of vaccinations was very high just after the initiation of the trial.

Interpretation: differential mortality impact of DTP and MV vaccinations

Compared with children who were MV-vaccinated at enrolment, the DTP-vaccinated but MV-unvaccinated children had twofold higher mortality. Using vaccination status as a predictor of subsequent vaccinations we found that the DTP0-2-vaccinated children had the highest mortality, whereas those who had already received MV and DTP3 had the lowest mortality, the difference being 2.5-fold.

Interestingly, the prevention of deaths attributed to measles by verbal autopsy explained little of this reduction in mortality. Similar observations have been made in previous studies from Guinea-Bissau (Aaby et al. 1996b), Senegal (Aaby et al. 1996a) and Bangladesh (Aaby et al. 2003a). MV-vaccinated children had worse WAZ z-score than the unvaccinated children, and it seems unlikely that the beneficial effect of MV is merely owing to a positive selection bias. As supported by previous studies (Aaby et al. 1995), MV may have a beneficial effect which is not explained by the prevention of measles infection.

On the other hand, being only DTP-vaccinated at enrolment was associated with higher mortality than the rates of both unvaccinated and fully vaccinated children. These children were likely to receive both MV and DTP during follow-up and usually administered simultaneously. Previous research has suggested that children who had had simultaneous administration of MV and DTP had higher subsequent mortality than children who had received MV on its own as the most recent vaccine (Aaby & Benn 2009b). Also, in a small randomised study, girls randomised to receive MV and DTP simultaneously had significantly poorer growth than children randomised to receive MV only (Agergaard et al. 2011). In the present study, the small group of DTP3-vaccinated children who were likely to receive only MV during follow-up also had low mortality. Control for baseline nutritional status did not substantially reduce the mortality differences between the various vaccination groups. The possibility that additional DTP vaccinations during follow-up may have had a negative effect was strengthened by the fact that children with incomplete DTP doses tended to have higher mortality than children with complete DTP doses at enrolment.

Given the uncertainty regarding who received additional vaccinations during follow-up, the Navrongo data on the NSE of vaccines should be interpreted with caution. Furthermore, selection biases may have explained some of the mortality differentials, but are unlikely to explain all of them because the mortality effect of different vaccines went in opposite directions. Specifically, the Navrongo data set documented similar differential effects of DTP and MV as have been reported in several previous studies (Velema et al. 1991Aaby et al. 1995, 2002, 2006c, 2012; Veirum et al. 2005).

Following studies suggesting a possible negative effect of DTP on survival (Aaby et al. 1995; Kristensen et al. 2000), several WHO-sponsored studies from Africa (Nyarko et al. 2001; Vaugelade et al. 2004; Elguero et al. 2005) have argued that DTP had strong beneficial effects, including a study from Navrongo based on data collected in 1996–2000 (Nyarko et al. 2001). However, these studies had survival bias, with the information on vaccinations received being better for children who survived (Aaby et al. 2007; Fine & Smith 2007; Jensen et al. 2007) because health cards for dead children were often not inspected for vaccination information after the death of a child. The family often destroys the card after the child’s death making vaccine status information for the dead children incomplete. With survival bias, vaccination will automatically be associated with a strong beneficial effect.

These data were collected 20 years ago when most children received vaccinations out-of-sequence and most had DTP as the last vaccination, and it could be speculated whether they had any relevance in the current situation. However, it needs to be emphasised that there are still many children getting vaccines out-of-sequence in rural Africa, particularly in states with less well-organised EPI (Aaby & Benn 2009b). Furthermore, the problem of having DTP as the last vaccination may become more common again. WHO’s SAGE (the Strategic Advisory Group of Experts on immunisation) has recently recommended a booster dose of pertussis vaccine given in the second year of life (WHO 2010), which would again make DTP the last vaccination through most of childhood.

Differences between sexes

In most studies of NSE of vaccines, we have emphasised that they tend to differ for boys and girls (Aaby et al. 1995, 2002, 2004b, 2006a,c, 2007, 2012; Benn et al. 2009). The data set in this study was too small to have been able to detect anything but very large differences. There may be other reasons that the sex-differential pattern was less clear in the present study. Firstly, because of the infrequent collection of vaccination status data, we could not evaluate the NSE of the last known vaccine that the child received with certainty. During follow-up, the children may have received different vaccines with opposite sex-differential effects, for example, MV after DTP or DTP after MV. Secondly, the majority of children in Navrongo received a live and inactivated vaccine simultaneously, that is, BCG and DTP or DTP and MV together. We have previously reported that a combined live and inactivated vaccine may be better for girls than for boys (Aaby et al. 2004a, 2009a).

Conclusion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and subjects
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Appendix

The present study adds to the growing evidence that the impact of routine vaccinations cannot be explained merely by prevention of targeted diseases or selection biases (Aaby et al. 2007) and that DTP and MV may have different effects for child survival (Velema et al. 1991; Kristensen et al. 2000; Aaby et al. 2002, 2004a, 2006a,c, 2009a, 2012; Veirum et al. 2005; Benn et al. 2009). Randomised controlled trials have documented that BCG and MV have non-specific beneficial effects (Aaby et al. 2010, 2011; Roth et al. 2010). There is a need for others to assess the impact of routine vaccinations in further randomised trials measuring mortality to improve the vaccination policy currently implemented in high-mortality low-income countries.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and subjects
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Appendix

Funding was provided by grants from DANIDA and European Union FP7 support for OPTIMUNISE. CB is funded through an ERC Starting Grant.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and subjects
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Appendix
  • Aaby P & Benn CS (2009b) Assessment of childhood immunisation coverage?Lancet373, 1428.
  • Aaby P, Samb B, Simondon F, Coll Seck AM, Knudsen K & Whittle H (1995) Non-specific beneficial effect of measles immunisation: analysis of mortality studies from developing countries. British Medical Journal311, 481485.
  • Aaby P, Samb B, Andersen M & Simondon F (1996a) No long-term excess mortality after measles infection: a community study from Senegal. American Journal of Epidemiology143, 10351041.
  • Aaby P, Lisse I, Mølbak K, Knudsen K & Whittle H (1996b) No persistent T lymphocyte immunosuppression or increased mortality after measles infection: a community study from Guinea-Bissau. The Pediatric Infectious Disease Journal15, 3944.
  • Aaby P, Garly ML, Balé C, Martins C, Lisse I & Jensen H (2002) Routine vaccinations and child survival in war situation with high mortality: effect of gender. Vaccine21, 1520.
  • Aaby P, Bhuyia A, Nahar L, Knudsen K, Francisco A & Strong M (2003a) The survival benefit of measles immunisation may not be explained entirely by the prevention of measles disease. International Journal of Epidemiology32, 106115.
  • Aaby P, Jensen H, Samb B et al. (2003b) Differences in female-male mortality after high-titre measles vaccine and association with subsequent vaccination with diphtheria-tetanus-pertussis and inactivated poliovirus: a re-analysis of the West African studies. Lancet361, 21832188.
  • Aaby P, Jensen H, Rodrigues A et al. (2004a) Divergent female-male mortality ratios associated with different routine vaccinations among female-male twin pairs. International Journal of Epidemiology33, 367373.
  • Aaby P, Jensen H, Gomes J, Fernandes M & Lisse IM (2004b) The introduction of diphtheria-tetanus-pertussis vaccine and child mortality in rural Guinea-Bissau: an observational study. International Journal of Epidemiology33, 374380.
  • Aaby P, Rodrigues A, Biai S et al. (2004c) Oral polio vaccination and low case fatality at the paediatric ward in Bissau, Guinea-Bissau. Vaccine22, 30143017.
  • Aaby P, Jensen H & Walraven G (2006a) Age-specific changes in female-male mortality ratio related to the pattern of vaccinations: an observational study from rural Gambia. Vaccine24, 47014708.
  • Aaby P, Ibrahim S, Libman M & Jensen H (2006b) The sequence of vaccinations and increased female mortality after high-titre measles vaccine: trials from rural Sudan and Kinshasa. Vaccine24, 27642771.
  • Aaby P, Vessari H & Nielsen J et al. (2006c) Non-specific and sex-differential effects of routine immunizations in rural Malawi. The Pediatric Infectious Disease Journal25, 721727.
  • Aaby P, Benn CS, Nielsen J, Lisse IM, Rodrigues A & Jensen H (2007) Estimating the effect of DTP vaccination on mortality in observational studies with incomplete vaccination data. Tropical Medicine and International Health12, 1524.
  • Aaby P, Benn CS, Nielsen J & Ravn H (2009a) Sex-differential non-specific effects of BCG and DTP in Cebu, The Philippines. International Journal of Epidemiology38, 320323.
  • Aaby P, Martins CL, Garly ML et al. (2010) Non-specific effects of standard measles vaccine at 4.5 and 9 months of age on childhood mortality: randomised controlled trial. BMJ341, c6495.
  • Aaby P, Roth A, Ravn H et al. (2011) Randomised trial of BCG vaccination at birth to low-birth-weight children: beneficial non-specific effects in the neonatal period?The Journal of Infectious Diseases204, 245252.
  • Aaby P, Benn CS, Nielsen J, Lisse IM, Rodrigues A & Ravn H (2012) Testing the hypothesis that diphtheria-tetanus-pertussis vaccine has negative non-specific and sex-differential effects on child survival in high-mortality countries. BMJ Open2, e000707.
  • Agergaard J, Nante E, Poulstrup G et al. (2011) Diphtheria-tetanus-pertussis vaccine administered simultaneously with measles vaccine is associated with increased morbidity and poor growth in girls. A randomised trial from Guinea-Bissau. Vaccine29, 487500.
  • Benn CS, Aaby P, Nielsen J, Binka FN & Ross DA (2009) Does vitamin A supplementation interact with routine vaccinations? An analysis of the Ghana vitamin A supplementation trial. The American Journal of Clinical Nutrition90, 629639.
  • Binka FN, Ross DA, Morris SS et al. (1995) Vitamin A supplementation and childhood malaria in northern Ghana. The American Journal of Clinical Nutrition61, 853859.
  • Elguero E, Simondon F, Simondon K & Vaugelade J (2005) Non-specific effects of vaccination on survival: a prospective study in Senegal. Tropical Medicine and International Health10, 956960.
  • Fine PEM & Smith PG (2007) ‘Non-specific effects of vaccines’– an important analytical insight, and a call for a workshop. Tropical Medicine and International Health12, 14.
  • Ghana VAST Study Team (1993) Vitamin A supplementation in northern Ghana: effects on clinic attendances, hospital admissions, and child mortality. Lancet342, 712. http://www.who.int/childgrowth/software/en
  • Jensen H, Benn CS, Nielsen J et al. (2007) Survival bias in observational studies of the impact of routine vaccinations on childhood survival. Tropical Medicine and International Health12, 514.
  • Kim-Farley RJ (1992) EPI for the 1990s. The Expanded Programme on Immunization Team. Vaccine10, 940948.
  • Kristensen I, Aaby P & Jensen H (2000) Routine vaccinations and child survival: follow up study in Guinea-Bissau, West Africa. BMJ321, 14351438.
  • Nyarko P, Pence B & Debpuur C (2001) Immunization Status and Child Survival in Rural Ghana. Population Council: Working papers No. 147, New York.
  • Ross DA, Kirkwood BR, Binka FN et al. (1995) Child morbidity and mortality following vitamin A supplementation in Ghana: time since dosing, number of doses and time of the year. American Journal of Public Health85, 12461251.
  • Roth A, Sodemann M, Jensen H et al. (2006a) Tuberculin reaction, BCG scar and lower female mortality. Epidemiology17, 562568.
  • Roth AE, Garly ML, Jensen H, Nielsen J & Aaby P (2006b) Bacille Calmette Guerin vaccination and infant mortality. Expert Review of Vaccines5, 277293.
  • Roth A, Benn CB, Ravn H et al. (2010) A randomised trial of the effect of revaccination with BCG in early childhood and mortality. BMJ340, c671.
  • Vaugelade J, Pinchinat S, Guielle G, Elguero E & Simondon F (2004) Lower mortality in vaccinated children: follow up study in Burkina Faso. BMJ329, 13091311.
  • Veirum JE, Sodemann M, Biai S et al. (2005) Diphtheria-tetanus-pertussis and measles vaccinations associated with divergent effects on female and male mortality at the paediatric ward in Bissau, Guinea-Bissau. Vaccine23, 11971203.
  • Velema JP, Alihonou EJ, Gandaho T & Hounye FH (1991) Childhood mortality among users and non- users of primary health care in a rural West African community. International Journal of Epidemiology20, 474479.
  • WHO (2010) Meeting of the Strategic Advisory Group of Experts on immunization, April 2010 – conclusions and recommendations. Weekly Epidemiological Record85, 197212.

Appendix

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and subjects
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Appendix
Ghana VAST Survival Study

The GhanaVAST Survival Study was conducted between 1989 and 1991 (Ghana VAST Study Team 1993; Binka et al. 1995; Ross et al. 1995; Benn et al. 2009). A census was carried out shortly before the trial started. The study area was exclusively rural and extended families lived in compounds. The area was divided in 185 clusters, which were randomised to vitamin A supplementation (VAS) or placebo within a double-blind trial. Children aged 6–90 months of age were enrolled into an open cohort and were visited at home every 4 months by trained fieldworkers, in a total of 7 rounds. They were dosed with vitamin A or placebo at each round. Over the 2 years, 21 906 children were enrolled in the trial by receiving at least one dose of either VAS or placebo; 13 462 being enrolled in Round 1. Children who moved out of the area were censored in the analysis at the time of moving; children who moved from a cluster with one treatment to a cluster with the opposite treatment were censored at the time of receiving the new treatment. Children who developed xerophthalmia were censored at the time of diagnosis and received special intensive vitamin A treatment. VAS was associated with a 19% (95%CI, 2–32%) reduction in mortality (Ghana VAST Study Team 1993).

Age determination

At enrolment, the date of birth of the child was determined from the health card if available. Otherwise, the date of birth was assessed by means of detailed local events calendars. If no exact date could be obtained, the date were set to day 15 of the month of birth. The date of birth may have been determined more precisely among children with a health card. For example, in Round 1, there were more children who had day 15 in a month as date of birth among children without a health card (3772/5066, 74.5%) than among those with a card (2764/6656, 41.5%) [Prevalence ratio 1.79 (1.69, 1.91)], and the month of birth was grouped close to the date of the enrolment month or 6 months earlier (P = 0.003), suggesting that the children had been reported to be for instance 1 year or 1½ year, rather than a year and some months. This may have made the children in the group with no health card appear to have been slightly younger than they truly were.

Information on vaccinations

Vaccination information was collected systematically for the children enrolled in Round 1. One year and 2 years later, in Rounds 4 and 7, vaccination information was collected again. Owing to an archiving error, dates of measles vaccination from Round 7 were not saved in the stored data files.

The collection of data on vaccinations was carried out de novo each time without taking previous information into consideration and only for children being alive at the visit. Thus, the vaccination status of children who died between vaccination visits was not updated. Comparing information from rounds 1 and 4 for the same children (see Appendix table), it is clear that some children made changes that should not have been possible, as all children in Round 4 should have had the same status or have received additional vaccinations since Round 1. However, 8% (122/1602) who originally had a vaccination card no longer had a card; 3% (30/931) of children who still had a health card and who were originally reported as having MV no longer had MV reported, and 3% (18/663) of children who still had a health card and who were originally fully DTP-vaccinated no longer had all three DTP vaccinations noted. Such errors presumably came from the field worker confusing the health cards of different children, or the field worker noting that the child had a card but forgetting to write the dates of some or all vaccinations.

The main analysis of the impact of vaccinations has been restricted to children enrolled in Round 1 for whom there was follow-up for the whole 2-year period. We have elsewhere analysed the impact of VAS on survival and the interactions with vaccinations (Benn et al. 2009). In the present analysis, we only included children who had received placebo as this group would not be subject to vaccine/vitamin A interactions.

Analysis of vaccinations

Vaccination information was only fully available with intervals of 1 year. Hence, it was difficult to analyse the mortality effect of the latest vaccines received until receiving the next vaccination as we have performed in previous studies. Furthermore, the incidence of vaccination was very high, particularly in the first months after enrolment, because of a major intensification of the vaccination programme before and immediately after 1990, the year when the Expanded Programme on Immunisation had targeted achieving 80% coverage of vaccination globally (Kim-Farley 1992). Therefore, a certain vaccination status in Round 1 or Round 4 not only reflected which vaccines the child had already received, but also which vaccines they were likely to receive in the coming months. Hence, any vaccination effects on mortality are likely to be a product of both vaccination status at enrolment and the vaccines likely to be received during follow-up. We have particularly focused on the impact on survival in the first 4 months when the information on immunisation status would be most valid. However, we have also reported the mortality effect for the full 2 years duration of the trial. Although some children who did not have a vaccination card may have had one but lost it, it is likely that most had never had a card and had never received any vaccination. Furthermore, data from subsequent rounds showed that the uptake of vaccination was low among children who initially had no health cards. For the present analyses, children who had no health card at Round 1 have been assumed to be ‘unvaccinated’.

Information from Rounds 1, 4 and 7 was used to calculate the proportion of children who received vaccinations in the years following enrolment. This proportion was estimated as the proportion of children who had received vaccination among children whose health card had been seen at enrolment and again 1 year or 2 years later, that is among surviving children. Censored children were excluded in calculating these proportions.

Among the vaccinated children, we have emphasised two aspects of vaccination status: first, children who had not yet received MV were likely to receive MV during follow-up; second, children who had not received all doses of DTP were likely to receive additional doses of DTP during follow-up. We have therefore distinguished between children who had received 0-2 doses (DTP0-2) or three or more doses of DTP (DTP3-4) at enrolment. Vaccination status at enrolment was categorised into six groups. Group 1 consisted of those with no health card and was considered ‘unvaccinated’. Groups 2–6 all had health cards at Round 1. Group 2 had not received DTP or MV (some of these children had, however, received BCG); Group 3 had received DTP1-2 but not MV; Group 4 had received DTP3-4, but not MV; Group 5 had received DTP1-2 and MV; and Group 6 had received DTP3-4 and MV (fully vaccinated).

Table Appendix Table1. Vaccination status at Rounds 1 and 4 for children who had a health card seen or had no card on both occasions. Navrongo, Ghana, 1989–1990
Vaccination status at enrolment (Round 1)Vaccination status at Round 4
1. No card2. Health card, no DTP, no MV3. Health card, DTP1-2, no MV4. Health card, DTP3-4, no MV5. Health card, MV, DTP 0-26. Health card, MV, DTP3-4Total
1No card6061720213286863
2Health card, no DTP, no MV2921704055152
3Health card, DTP1-2, no MV317561058204366
4Health card, DTP3-4, no MV3201007994
5Health card, MV, DTP0-235272142206394
6Health card, MV, DTP3-424321411542596
Total 72852923838311722465
Table Appendix Table2. Mortality rate (MR) per 1000 person-years (pyrs) and mortality rate ratio (MRR) for vaccinated compared with unvaccinated children, among children <36 months at enrolment into the placebo group
Group: Vaccination status at enrolment N Median weight-for-age z-score at enrolmentMR per 1000 years (deaths/pyrs) {measles deaths}Within 4 months follow-upWithin 24 months follow-up
Unadjusted MRR (95% CI) {MRR – without measles deaths)MRR* (95% CI) (Vaccinated/No heath card) {MRR – without measles deaths}MR per 1000 years (deaths/pyrs) {measles deaths}Unadjusted MRR (95% CI) {MRR – without measles deaths)MRR* (95% CI) (Vaccinated/No health card) {MRR – without measles deaths}
  1. *Adjusted for age, zone, WAZ and ownership of radio (N = 3082).

1. No card1093−1.7 [−2.6; −0.8] 80 (31/389) {2}1.0 (Reference)1.0 (Reference)51 (101/1998) {23}1.0 (Reference)1.0 (Reference)
2. Health card:  noDTP, noMV196−1.9 [−2.9; −1.0] 58 (4/69)0.72 (0.25, 2.05) {0.77 (0.27, 2.19)}0.65 (0.23, 1.84) {0.71 (0.25, 2.04)}30 (11/370) {2}0.59 (0.32, 1.10) {0.63 (0.31, 1.25)}0.53 (0.28, 0.99) {0.57 (0.28, 1.13)}
3. DTP1-2,  no MV469−2.0 [−3.0; −1.2]152 (25/164)1.90 (1.12, 3.22) {2.03 (1.19, 3.47)}1.79 (1.03, 3.11) {1.87 (1.07, 3.26)}54 (46/854) {6}1.07 (0.75, 1.51) {1.19 (0.82, 1.75)}0.93 (0.65, 1.33) {1.05 (0.71, 1.56)}
4. DTP3-4,  no MV109−1.8 [−3.0; −1.2] 52 (2/38)0.66 (0.16, 2.74) {0.70 (0.17, 2.94)}0.65 (0.15, 2.74) {0.68 (0.16, 2.90)}24 (5/207) {1}0.48 (0.20, 1.19) {0.50 (0.18, 1.36)}0.45 (0.18, 1.11) {0.47 (0.17, 1.30)}
5. MV, DTP0-2485−1.9 [−2.8; −1.1] 47 (8/171){2}0.58 (0.27, 1.27) {0.47 (0.19, 1.23)}0.74 (0.34, 1.65) {0.60 (0.24, 1.47)}26 (24/923) {3}0.52 (0.33, 0.81) {0.58 (0.36, 0.94)}0.58 (0.37, 0.91) {0.70 (0.42, 1.14)}
6. MV, DTP3-4730−1.8 [−2.6; −1.0] 35 (9/257)0.44 (0.21, 0.92) {0.47 (0.22, 0.98)}0.61 (0.28, 1.35) {0.67 (0.31, 1.48)}14 (20/1408)0.28 (0.18, 0.46) {0.36 (0.22, 0.60)}0.35 (0.21, 0.57) {0.47 (0.28, 0.79)}
Total3082−1.8 [−2.7; −1.0]73 (79/1090){4}  36 (207/5760){35}