The Introduction of Diphtheria-Tetanus-Pertussis and Oral Polio Vaccine
Among Young Infants in an Urban African Community: A
Søren Wengel Mogensen
, Andreas Andersen
, Amabelia Rodrigues
, Christine S Benn
, Peter Aaby
Bandim Health Project, Indepth Network, Apartado 861, Bissau, Guinea-Bissau
Research Centre for Vitamins and Vaccines (CVIVA), Bandim Health Project, Statens Serum Institute, Artillerivej 5, 2300 Copenhagen S, Denmark
OPEN, Institute of Clinical Research, University of Southern Denmark/Odense University Hospital, 5000 Odense C, Denmark
Received 4 June 2016
Received in revised form 21 January 2017
Accepted 29 January 2017
Available online 1 February 2017
Background: We examined the introduction of diphtheria-tetanus-pertussis (DTP) and oral polio vaccine (OPV)
in an urban community in Guinea-Bissau in the early 1980s.
Methods:The child population had been followed with 3-monthly nutritional weighing sessions since 1978.From
June 1981 DTP and OPV were offered from 3 months of age at these sessions. Due to the 3-monthly intervals be-
tween sessions, the children were allocated by birthday in a ‘natural experiment’to receive vaccinations early or
late between3 and 5 months of age. We included children who were b6 monthsof age when vaccinations started
and children born until the end of December 1983. We compared mortality between 3 and 5 months of age of
DTP-vaccinated and not-yet-DTP-vaccinated children in Cox proportional hazard models.
Results: Among 3–5-month-old children, having received DTP (±OPV) was associated with a mortality hazard
ratio (HR ) of 5.00 (95% CI 1.53–16.3) comparedwith not-yet-DTP-vaccinatedchildren. Differences in background
factors did not explain the effect. The negative effect was particularly strong for children who had received DTP-
only and no OPV (HR = 10.0 (2.61–38.6)). All-cause infant mortality after 3 months of age increased after the
introduction of these vaccines (HR = 2.12 (1.07–4.19)).
Conclusion: DTP was associated with increased mortality; OPV may modify the effect of DTP.
© 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
Non-speciﬁc effects of vaccines
Oral polio vaccine
Individually randomized studies to measure impact on child survival
of different vaccines were not conducted when the Expanded Program
on Immunization (EPI) was introduced in low-income countries in the
1970s. The disease-protective effects were well documented, so the
main issue was at which age to introduce the vaccine most effectively
(The Expanded Programme on Immunization, 1982). Except for mea-
sles vaccine (MV), surprisingly few studies examined the introduction
of vaccines and their impact on child survival (Aaby et al., 1983,
2003a; Holt et al., 1990; The Kasongo Project Team, 1981). One trial of
measles-vaccinated and measles-unvaccinated communities in Congo
showed a larger than expected reduction in child mortality (Aaby et
al., 1981); this observation was subsequently corroborated by commu-
nity “trials”and before-after studies in several countries (Aaby et al.
1984, 1993, 2003a; Holt et al., 1990; Kapoor and Reddaiah, 1991).
Hence, a vaccine may have non-speciﬁc effects (NSEs) on susceptibility
to other infections (Aaby et al., 1995). WHO's Strategic Advisory Group
of Experts on Immunization (SAGE) recently reviewed the potential
NSEs of BCG, diphtheria-tetanus-pertussis (DTP) and MV and recom-
mended further research (Higgins et al., 2014; Strategic Advisory
Group of experts on Immunization, 2014).
Though protective against the target diseases, DTP may increase sus-
ceptibility to unrelated infections (Aaby et al., 2003b, 2004a, 2012)
(Appendix A). The SAGE review noticed that the majority of studies
found a detrimental effect of DTP (Higgins et al., 2014). However,
SAGE considered the evidence inconsistent because two studies report-
ed beneﬁcial effects (Higgins et al., 2014) and that most studies
underestimated the beneﬁt of DTP because studies were conducted in
situations with herd immunity. Furthermore, all studies gave DTP and
OPV together, making it impossible to separate effects of DTP and OPV
(SAGE non-speciﬁc effects of vaccines Working Group, 2014).
On the other hand, the “unvaccinated”children in these studies have
usually been frail children too sick or malnourish to get vaccinated, and
the studies may therefore have underestimated the negative effect of
DTP. We therefore examined what happened when DTP and OPV
were ﬁrst introduced, but not always given together, in 1981–1983 in
EBioMedicine 17 (2017) 192–198
⁎Corresponding author at: Bandim Health Project, Statens Serum Institute, Artillerivej
5, 2300 Copenhagen S, Denmark.
E-mail address: firstname.lastname@example.org (P. Aaby).
2352-3964/© 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Contents lists available at ScienceDirect
journal homepage: www.ebiomedicine.com
the capital of Guinea-Bissau. In this situation the children were allocated
by birthday to receive vaccines early or late and the “unvaccinated”
were therefore not frail children.
Bandim Health Project (BHP) has followed an urban community
with a demographic surveillance system since December 1978, and
took part in the introduction of vaccines well before a full-ﬂedged na-
tional program was implemented with UNICEF support in 1986 (Aaby
et al., 1984, 2004a).
2.2. Demographic Surveillance
In 1978–1979, under-ﬁve mortality was nearly 500/1000. Since mal-
nutrition was assumed to be the main cause, a study was initiated to de-
termine why children were malnourished (Aaby et al., 1983). However,
severe malnutrition was not evident, and to understand the high mor-
tality we started a health and demographic surveillance system
(HDSS). The area was mapped and a census conducted. Four health
workers were employed to identify pregnant women, encourage
women to attend ante-natal clinics, and to follow children with anthro-
pometric measurements to assess growth patterns and detect malnour-
ished children. Each health worker followed a population of 1500–2000
individuals. The health workers were supervised by anexpatriate nurse.
For each sub-district in Bandim, the responsible health worker kept
a list of children under three years of age. BHP had no computerized sur-
veillance system until 1990 but kept an A5 card (“BHP card”) for each
child, where weights and vaccination dates were noted. The child's
growth card was kept by the mother.
The Bandim population was very mobile. It was important to main-
tain contact with the natal village for ceremonial purposes and to secure
rice. Furthermore, mothers were not supposed to have sexual relations
during breastfeeding (Jakobsen et al., 2004). Breastfeeding was
prolonged in Guinea-Bissau. Thus, many women stayed in the rural
areas with their natal family while breastfeeding. These cultural
traditions introduced variability in the participation in weighing and
We arranged quarterly weighing sessions in each sub-district. The
responsible health worker advisedmothers the day before a community
weighing. The following morning, the weight was measured and noted
on the child's growth card and the BHPcard. When the World Food Pro-
gram provided supplementary feeding this was given to families with
There was no community vaccination program in 1981 except that
we had organized a few measles vaccination campaigns (Aaby et al.,
1984). Mothers could take their children to the Mother and Child Health
Program in town. However, this clinic was mainly attended by the
urban elite. Few children were vaccinated before BHP organized vacci-
nation sessions (Table 1).
In June 1981, BHP started to provide vaccinations at the quarterly
weighing sessions. A health center nurse accompanied the weighing
team and vaccinated eligible children. DTP and OPV were provided
from 3 months and MV from 9 months of age. OPV-at-birth was not
given then.The three DTP and OPV doses could be given with an interval
of one month but since we only arranged weighing every three months,
most children had longer intervals between doses. DTP was adminis-
tered intramuscularly and OPV as an oral drop. When both vaccines
were administered at the same session OPV was usually given ﬁrst
and then DTP; the children would usually start crying after DTP due to
the pain of the injection and it would therefore have complicated the
administration of OPV to give DTP ﬁrst. There were several periods
where either OPV or DTP was missing (Fig. 1). BCG was rarely provided
at the weighing sessions since most nurses were not trained to admin-
ister intra-dermal vaccination. A total of 269 children may have been
BCG vaccinated as they had a vaccination date on their card (N =
192) or were noted to have received BCG but no date given (N = 77).
The expatriate nurse sometimes organized additional vaccination
sessions in which the children were not weighed. During these sessions,
Median age of vaccination and coverage for BCG, DTP and OPV of study cohort.
1980 1981 1982 1983 1981–1983
Median age in days (N vaccines)
BCG 9 (4) 48.5 (50) 34 (46) 25 (68) 33 (164)
DTP1 97 (12) 127 (147) 121 (164) 117 (278) 121 (589)
OPV1 98 (12) 118 (185) 121.5 (170) 117 (225) 118 (580)
MV 181 (5) 141 (53) 157 (2) 110 (1) 141.5 (56)
Coverage at 6 months of age
BCG 1.7% (5/289) 3.5% (12/342) 23.7% (72/304) 17.4% (57/327) 14.5% (141/973)
DTP1 4.2% (12/289) 31.3% (107/342) 61.2% (186/304) 73.1% (239/327) 54.7% (532/973)
DTP3 2.4% (7/289) 0.9% (3/342) 4.3% (13/304) 4.0% (13/327) 3.0% (29/973)
OPV1 4.2% (12/289) 43.0% (147/342) 62.5% (190/304) 69.7% (228/327) 58.1% (565/973)
OPV3 2.4% (7/289) 2.0% (7/342) 4.3% (13/304) 4.0% (13/327) 3.4% (33/973)
MV 2.8% (8/289) 15.2% (52/342) 0.7% (2/304) 0% (0/327) 5.5% (54/973)
Coverage at one year of age
BCG 2.6% (3/116) 2.4% (7/294) 15.4% (51/332) 17.4% (46/264) 11.7% (104/890)
DTP1 2.6% (3/116) 32.7% (96/294) 71.1% (236/332) 83.0% (219/264) 61.9% (551/890)
DTP3 2.6% (3/116) 4.4% (13/294) 18.4% (61/332) 43.2% (114/264) 21.1% (188/890)
OPV1 2.6% (3/116) 37.4% (110/294) 77.4% (257/332) 84.8% (224/264) 66.4% (591/890)
OPV3 2.6% (3/116) 12.2% (36/294) 32.5% (108/332) 44.3% (117/264) 29.3% (261/890)
MV 15.5% (18/116) 68.0% (200/294) 34.0% (113/332) 51.1% (135/264) 50.3% (448/890)
Notes: The inclusion criteria for thecohort in Table 1 are the same as for our study cohort: weight examination after 15 days of age and contribute time between 91 and 183 daysof age.
Median age: ‘year’means the year the vaccination was given, and median age is the median age at time of vaccination with a given vaccine among children vaccinated before turning
6 months. E.g. the 4 BCG vaccines in the 1980 column were given in 1980 to children with a median age of 9 days.
Coverage: ‘year’meansthe year when thechild turned exactly1 year (or 6 months)old and coveragewas assessed. Onlychildren survivingto 1 year (or 6 months)of age were assessedfor
coverage. Children turning 1 year in 1984 were thus not presented in the table.
193S.W. Mogensen et al. / EBioMedicine 17 (2017) 192–198
vaccinations were noted on the BHP cards. Both nurses and mothers
thought that sick children should not be vaccinated; the BHP card
often indicated that the child was ‘sick’,‘malnourished’or ‘orphan’as
an explanation of why an age-eligible child had not been vaccinated.
2.5. Data Control
When a computerized system became available in 1990–1991,
weights and vaccinations from the BHP cards were entered. For the
present analysis, all information on dates of visit, weights and vaccina-
tion dates was checked against the original cards. A few cards were
not available or could no longer be found (Fig. 2).
2.6. The Study Cohort
We included children born from December 3, 1980 as they would
become eligible for vaccination before 6 months of age (Fig. 2). Few chil-
dren were vaccinated with BCG (Table 1). Children who travelled and
never attended any session were not included in the ‘unvaccinated’
group. Children weighed within a fortnight of their birth to obtain a
birth weight were only included if they took part in a subsequent com-
munity weighing session. Furthermore, we excluded orphans since they
were not breastfed and were likely to have different care. The cohort is
depicted in Supplementary Fig. 1.
2.7. Natural Experiment for 3–5-month-old Children
Though not individually randomized, the present study is a natural
experiment with limited bias ingroup allocation:With 3-monthlyinter-
vals between weighing sessions, children were allocated by their birth-
day to receive their ﬁrst vaccinations early or late between 3 and
5monthsofage(Fig. 3). We therefore compared 3–5-month-old chil-
dren who had received DTP (±OPV) vaccinations early with children
who had not yet received these vaccinations. Since there were no
healthy “unvaccinated”children after 6 months of age unless they had
travelled, we censored follow-up of all children at 6 months of age
Sick children were not vaccinated, in the main analysis we therefore
censored ‘unvaccinated’children who attended a weighing session but
did not receive a vaccination (Fig. 3). Since the censoring of sick children
could have introduced a bias, we also conducted an intention-to-treat
analysis in which the censored children were transferred to the DTP
group. Hence, in this analysis we compared the mortality of the
intended-DTP-vaccinated group and the not yet DTP-vaccinated group.
Children were included from 91 days of age if they had been exam-
ined in a weighing session before 91 days; if they were only seen in a
weighing session after 3 months of age they were only included from
the day seen. DTP was not administered elsewhere and the follow-up
time of children was therefore counted as DTP-unvaccinated time in
the survival analysis until BHP provided the vaccine. Time as DTP-un-
vaccinated also came fromchildren who did not turn up at the weighing
sessions between 3 and 5 months of age but had been seen before
3 months of age and therefore were part of the community cohort
(Fig. 3). Hence, the DTP-vaccinated and DTP-unvaccinated children
were all children from the same cohort of children born in Bandim
and their allocation depended on the timing of their birth date, the
timing of the weighing sessions and their travelling pattern. We
Fig. 1. Each vaccination of the speciﬁed type is plotted according age of the recipient and date of vaccination.
Fig. 2. Flowchart of studypopulation and children included in the analyses. Notes: DOB =
date of birth;  indicates the number of deaths before 6 months of age in the group.
194 S.W. Mogensen et al. / EBioMedicine 17 (2017) 192–198
compared the background factors for the children who were DTP vacci-
nated, attended a weighing session between 3 and 5 months but were
not vaccinated and those who did not attend a weighing session
We also examined the mortality of children who due to logistic rea-
sons had received DTP-only. Absences and travelling patterns are un-
likely to differ between children who at their ﬁrst vaccination had
received DTP1 + OPV versus DTP1-only; thesetwo groups were equally
likely to receive subsequent vaccinationsboth with respect to timing of
subsequent vaccinations and coverage (data available on request).
2.8. Statistical Methods
First possible enrolment date was June 2, 1981, when DTP and OPV
vaccinations were introduced. Different vaccination groups were com-
pared using a Cox proportional hazard model with age as underlying
Children were classiﬁed according to their most recent vaccination
(Supplementary Table 1). We ignored BCG vaccinations in the main
analysis because we gave few BCG vaccinations (Table 1) and some chil-
dren had received BCG at the maternity ward without proper documen-
tation as some children had a BCG scar but no vaccination card. To avoid
survival bias, we used a landmark approach (Jensen et al., 2007); hence,
a child's vaccination status was only updated from the day the informa-
tion was collected. Due to the additional vaccination sessions organized
by the expatriate nurse some “unvaccinated”children received a vac-
cine before the weighing session where they changed status to “vacci-
nated”; it is noted in the footnote to Table 3 how many had received
such vaccinations. As a sensitivity analysis we also did an analysis in-
cluding the additional vaccination sessions as landmarks. For the re-
mainder of this paper, we will refer to these landmarks as vaccination-
The WHO z-score for weight-for-age was used to assess nutritional
status. Control for sub-district, ethnic group and twinning did not
change the results (data not shown). There was no obvious clustering
Fig. 3. Natural experiment studydesign. Note: Children were weighed everythird month. After3 months of age theyreceived DTP and OPVon weighing daysif they were healthy.Children
who attendedbut were not vaccinated at a weighing session after 3 months of age were censored in the survival analysis comparing DTP-vaccinated and unvaccinated children.
Background factors children in the main analysis of vaccination and mortality between 3 and 5 months of age.
Attended weighing session at
3–5 months, not vaccina ted
Did not attend weighing
session at 3–5 months
Number 662 186 209
Male sex 52.1% 53.2% 54.1%
Twin 2.7% 2.2% 2.9%
Birth weight (SD) 3.23 (0.025) 3.28 (0.061) 3.22 (0.051)
•Pepel 46.8% 54.8% 45.0%
•Balanta 11.8% 13.4% 16.3%
•Other ethnic groups 41.4% 31.7% 38.8%
Mean weight-for-age z-score (SD) at examination before 3 months of age −0.30 (0.037) −0.34 (0.084) −0.43 (0.066)
Follow-up time (person-years) between 3 and 5 months;
[Median number of days of follow]
All time 135.5  36.8  47.4 
73.3 1.8 2.0
62.2 35.1 45.4
Mean number (SD) of weighing sessions per year between 6 and 11 months
2.7 (0.03) 2.2 (0.07) 1.6 (0.08)
195S.W. Mogensen et al. / EBioMedicine 17 (2017) 192–198
of deaths and control for season and calendar time did not change esti-
mates (data not shown).
There were 18 deaths between 3 and 5 months of age: 3 had cough
and respiratory infections as the main symptom, 3 had fever (presumed
malaria), 2 were due to diarrhea, 5 had diarrhea and vomiting, 1 was a
sudden death, and 4 had no information on cause.
The study of nutritional status was planned by SAREC (Swedish
Agency for Research Collaboration with Developing Countries) and the
Ministry of Health in Guinea-Bissau.
Of 1356 children registered in Bandim and followed to 3 months of
age (Fig. 2), 286 were never weighed, had no card or their card was
lost. An additional 13 children had inconsistent information, vaccinations
marked with a cross but without dates or were orphans. Hence, 1057 chil-
dren were included in the study cohort. The median ages for DTP1 and
OPV1 were 121 and 118 days, respectively (Table 1). The vaccination cov-
erage at 6 months of age was 55% for DTP1; 3% got DTP3 (Table 1). Cover-
age for MV was only 6%. Of the DTP1, OPV1 and MV vaccinations noted on
the BHP card 90–95% had been administered by the BHP.
For children examined after 91 days, a one-unit increase in w/a z-
score was associated with an odds ratio of 1.07 (0.93–1.24) for receiving
a vaccination at that weighing session.
3.1. Natural Experiment with 3–5-month-old Children
There were no marked differences in background factors for the three
groups of children who were DTP vaccinated at 3–5monthsofage,those
who attended a weighing session but were not vaccinated, and those who
did not attend a weighing session at 3–5monthsofage(Table 2). Birth
weight was similar in the three groups. Weight-for-age z-score before
3 months of age did not differ for the three groups (Table 2). Those who
did not attend a weighing session at 3–5 months of age were signiﬁcantly
less likely to attend later weighing sessions during infancy, the mean
number of visits being lower for those not attending than for those
being DTP-vaccinated (p b0.001) (Table 2); hence, they are likely to
have travelled more than those who were DTP-vaccinated.
In the main experiment depicted in Fig. 3, DTP vaccination (±OPV)
compared with ‘DTP-unvaccinated’was associated with a HR of 5.00
(1.53–16.3) (Table 3); the HR was 9.98 (0.81–123) for girls and 3.93
(1.01–15.3) for boys. If we also included vaccinations given on vaccina-
tions-days-without-weighing in the landmark analysis, DTP (± OPV)
compared with unvaccinated was associated with a HR of 3.90 (1.20–
12.3). When DTP had been given alone without OPV the HR was 10.0
(2.61–38.6) (Table 3). The difference between DTP-only children and
DTP-plus-OPV does not reﬂect differences in follow-up and other vacci-
nations since the timeto DTP2 and prevalence of DTP2 was the same for
DTP-only and DTP-plus-OPV vaccinated children (data not shown). If
we excluded the 269 children who may have been BCG vaccinated re-
sults were similar (Supplementary Table 2).
If the analysis was conducted as an intention-to-treat analysis in
which the children weighed but not vaccinated were not censored but
transferred to the DTP group, the intended-DTP-vaccinated group had
a HR of 3.92 (1.20–12.8) compared with the not-yet vaccinated group
(Supplementary Table 3).
3.2. Secondary Analyses
Since the introduction of DTP and OPV apparently was associated
with increased mortality, we examined what happened to infant mor-
tality from 3 to 12 months of age after the introduction of these vac-
cines. The mortality rate for all 3–11 months old children increased 2-
fold (HR = 2.12 (1.07–4.19)) from 1980, before vaccinations, to
1982–1983, after the introduction of DTP and OPV (Table 4).
4.1. Main Observations
DTP vaccinations were associated with increased infant mortality even
though there was no vaccine-induced herd immunity. When unvaccinated
controls were normal children who had not yet been eligible for vaccina-
tion, mortality was 5 times higher for DTP-vaccinated children. Co-adminis-
tration of OPV with DTP may have reduced the negative effects of DTP.
4.2. Strength and Weaknesses
The present analysisassessed DTP and child survival in a “natural ex-
periment”in which the children were allocated by the timing of their
birth and community weighing sessions and the group allocation was
therefore not inﬂuenced by the usual selectionbiases to the same extent
as most other studies of DTP (Aaby et al., 2016). To assure that the cen-
soring from the main analysis of children who were not vaccinated had
not produced the unexpected strong result we made an intention-to-
treat analysis but this did not change the result. If anything the un-
vaccinated children had slightly worse nutritional status before
3 months of age than the children who were subsequently DTP vac-
cinated(p=0.09)(Table 2); the unvaccinated children travelled
more than the DTP vaccinated children. These biases would tend to
favor rather than increase mortality in the DTP group and the
Mortality rate and hazard rate (HR) for children from 3 months of age until ﬁrst examination without vaccination or 6 months of age. Natural experiment.
3–5 months Mortality rate (deaths/person-years) HR (95% CI) for DTP vs unvaccinated
(N = 651)
4.5 (5/111.4) DTP (± OPV) (N = 462) 17.4 (11/63.1) 5.00 (1.53–16.3)
DTP only (N = 101) 35.2 (5/14.2) 10.0 (2.61–38.6)
DTP + OPV (N = 361) 12.3 (6/48.9) 3.52 (0.96–12.9)
(N = 313)
1.9 (1/51.9) DTP (± OPV) (N = 222) 13.3 (4/30.1) 9.98 (0.81–123.0)
DTP only (N = 44) 16.2 (1/6.2) 12.0 (0.56–257.2)
DTP + OPV (N = 178) 12.5 (3/23.9) 9.50 (0.73–124.0)
(N = 338)
6.7 (4/59.5) DTP (± OPV) (N = 240) 21.2 (7/33.0) 3.93 (1.01–15.3)
DTP only (N = 57) 49.8 (4/8.0) 8.93 (2.01–39.7)
DTP + OPV (N = 183) 12.0 (3/24.9) 2.21 (0.44–11.0)
Notes:There were no deaths due accidents in thisage group. BCG is disregarded in the analysis.Hence, the unvaccinated children have not received DTP, OPVor MV but may have received
BCG. Of the 651 unvaccinated children, 219 received DTP and/orOPV before their ﬁrst weighing examination. These children counted as ‘unvaccinated’until their ﬁrst weighing exami-
nation. Of the 462 children who received DTP (±OPV), 177 receivedan additional DTP or OPV before 6 monthsof age. The OPV-only is not presented in the table because there were no
deaths and very little follow-up time in this age group.
196 S.W. Mogensen et al. / EBioMedicine 17 (2017) 192–198
estimates from the natural experiment may therefore still be
The estimated effects of DTP andOPV are unlikely tohave been inﬂu-
enced by other vaccinations since very few had received other vaccines;
if the children who may have received BCG were censored in the analy-
sis the result was essentially the same (Supplementary Table 2).
The 3-monthly community examinations assured that we had fol-
low-up information for all children and relatively accurate information
on the time of death. Some children were excluded because a BHP
card could not be found and we did not know whether they had been
vaccinated or were travelling. Most likely, BHP cards may never have
been made because the child was not coming for examination, or the
card may have disappeared at community examinations, at the later
handling of BHP cards by ﬁeld workers or data entry clerks, or due to
mice. However, the few missing cards are unlikely to have affected the
main analysis as the mortality rate in this group was similar to the gen-
eral mortality rate (Fig. 2).
To assure comparability of vaccinated and unvaccinated groups, also
with respect to travelling, we included only children who had been
weighed in Bandim in connection with the 3-monthly community ex-
aminations. This meant that children who mostly stayed outside the
area were not included in the analysis; these children had no access to
community vaccinations and they lived elsewhere where the mortality
risk might have been quite different, e.g. due to a higher risk of malaria
The present study was not a planned trial. The study would have
been a cleaner natural experiment if vaccinations had only been admin-
istered at the weighing sessions. However, the expatriate nurse did or-
ganize additional vaccinations and some ‘unvaccinated’children had
therefore already received a vaccination before coming for a weighing
session. These ‘misclassiﬁcations’do not explain the increased mortality
in the DTP group. The estimate for DTP-vaccinated (±OPV) compared
with DTP-unvaccinated children was 4-fold higher mortality when we
included these additional landmarks in the analysis.
4.3. Comparison with Previous Studies of DTP and OPV
There is only one other study of the introduction of DTP. In rural
Guinea-Bissau, DTP (±OPV) was associated with 2-fold higher mortal-
ity (Aaby et al., 2004a). All studies that documented vaccination status
and followed children prospectively indicate that DTP has negative ef-
fects; a meta-analysis of the eight studies found 2-fold higher mortality
for DTP-vaccinated compared with DTP-unvaccinated, mostlyBCG-vac-
cinated controls (Aaby et al., 2016)(Appendix A).
The negative effect of DTP was much worse in this natural experi-
ment than has been reported in previous studies of DTP. This is presum-
ably due to the “unvaccinated”control children in previous studies
having been a frail subgroup too frail to get vaccinated. Previous studies
have not been able to compare DTP-vaccinated children with “normal”
controls. Hence, most previous studies have probably underestimated
the negative effect of DTP.
The potentially differential effects of DTP and OPV have only been
examined in few studies. However, we have recently been able to doc-
ument marked beneﬁcial NSEs of OPV. In an RCT, OPV at birth (OPV0)
reduced infant mortality by 32% (0–57%) before the children received
campaign-OPV (Lund et al., 2015). In Bissau campaign-OPV reduced
the mortality rate by 19% (5–32%) (submitted). When DTP was missing
for several months in Bissau, we showed that the all-cause case-fatality
at the pediatric ward was 3-fold lower if the children had OPV-only as
their most recent vaccination rather than the recommended combina-
tion of DTP and OPV (Aaby et al., 2004b). Thus, OPV may have modiﬁed
the negative effect of DTP.
This pattern was also seen when DTP was ﬁrst introduced in the
rural areas of Guinea-Bissau in 1984 (Aaby et al., 2004a). OPV was not
used the ﬁrst year and the HR for DTP versus unvaccinated was 5.00
(0.63–39.7). In the period from 1985 to 1987, when DTP and OPV
were nearly always administered together, the MRR was 1.90 (0.91–
3.97). In the present study, the hazard ratio was 10.0 (2.61–38.6) for
DTP-only but 3.52 (0.96–12.9) for children who received DTP and OPV
simultaneously (Table 3). Based on these two studies of the introduc-
tion of DTP, the HR compared with DTP-unvaccinated children was sig-
niﬁcantly different for children who had received DTP-only (HR = 8.14
(2.63–15.2)) and for children who received both DTP and OPV (HR =
2.21 (1.16–4.19)) (test of interaction, p = 0.049). Hence, simultaneous
administration of DTP and OPV may have alleviated the negative non-
speciﬁc effect of DTP.
DTP was associated with 5-fold higher mortality than being unvacci-
nated. No prospective study has shown beneﬁcial survival effects of
DTP. Unfortunately, DTP is the most widely used vaccine, and the pro-
portion who receives DTP3 is used globally as anindicator of the perfor-
mance of national vaccination programs.
It should be of concern that the effect of routine vaccinations on all-
cause mortality was not tested in randomized trials. All currently avail-
able evidence suggests that DTP vaccine may kill more children from
other causes than it saves from diphtheria, tetanus or pertussis. Though
a vaccine protects children against the target disease it may simulta-
neously increase susceptibility to unrelated infections.
The recently published SAGE review called for randomized trials of
DTP (Higgins et al., 2014). However, at the same time the IVIR-AC com-
mittee to which SAGE delegated the follow-up studies of the NSEs of
vaccines has indicated that it will not be possible to examine the effect
of DTP in an unbiased way. If that decision by IVIR-AC remains unchal-
lenged, the present study may remain the closest we will ever come
to a RCT of the NSEs of DTP.
The presentstudy and cleaning of the original data was supported by
a common grant from DANIDA and the Novo Nordisk Foundation (FU-
11-551). The work on non-speciﬁc effects of vaccines has been support-
ed by the Danish Council for Development Research, Ministry of Foreign
Affairs, Denmark [grant number 104.Dan.8.f.], Novo Nordisk Foundation
and European Union FP7 support for OPTIMUNISE (grant: Health-F3-
2011-261375). CSB held a starting grant from the ERC (ERC-2009-StG-
243149). CVIVA is supported by a grant from the Danish National Re-
search Foundation (DNRF108). PA held a research professorship grant
from the Novo Nordisk Foundation.
Mortality rates (deaths/100 person-years) between 3 and 11 months of age by study year.
Mortality rate 1980 1981 1982 1983
HR (95% CI) for 1982–1983 versus
(N = 547)
(N = 678)
(N = 632)
(N = 638)
Notes: Event recorded as accidents were not removed from this analysis.
197S.W. Mogensen et al. / EBioMedicine 17 (2017) 192–198
Conﬂict of Interest
Nothing to declare
CSB and PA proposed the study. PA collected the original data. AR is
responsible for the demographic surveillance system. SWM and PA
cleaned the data. SWM and AA conducted the statistical analyses. The
ﬁrst draft was writtenby PA; all authors contributed to the ﬁnal version
of the paper. PA and SWM will act as guarantors of the study.
The funding agencies had no role in the study design, data collection,
data analysis, data interpretation, or the writing of the report.
Through request to the authors
Appendix A. The DTP Controversy
The issue of DTP vaccination and child mortality in high mortality
areas was raised 15 years ago when a study from rural Guinea-Bissau
showed 1.84-fold higher mortality for children who had received
DTP1 vaccination (Aaby et al., 2016; Kristensen et al., 2000). All subse-
quent prospective studies have supported a negative effect (Aaby et
al., 2016). Furthermore, DTP may have a negative effect when given si-
multaneously with or after MV (Aaby at el., 2003b, 2012). For example,
the negative effect of high-titer measles vaccination (HTMV) in girls,
which led to the global withdrawal of HTMV, was due to DTP being ad-
ministered after MV because HTMV had been given early at 4–5months
of age (Aaby et al., 2003b).
DTP has not been shown to have beneﬁcial effects in RCTs or natural
experiments. The current policy for DTP has only been examined by
reanalyses of existing data sets collected for other purposes. All such
studies have had methodological problems related to different forms
of frailty and survival bias (Aaby et al., 2012). These studies have up-
dated follow-up time for DTP-vaccinated children who survived but
children who died without their vaccination status being documented
were classiﬁed as “unvaccinated”. Such procedures give a misleading
high mortality rate in the unvaccinated group, and the comparison of
DTP-vaccinated survivors and “unvaccinated”children will therefore
give a beneﬁcial estimate for DTP (Aaby et al., 2016). If the mortality
rate of unvaccinated children is unnaturally increased, the HR of unvac-
cinated children versus children who have received at least one vaccine
may indicate how much bias there might be in the study, and we have
called this HR the “bias-index”. All studies with prospective follow-up
have had a bias index below 2.0 (Aaby et al., 2016); in the present
study the bias index was 0.41 (0.15–1.15) in the 3–5 months age
group (Supplementary Table 2). In studies with survival bias and unnat-
urally high mortality in the unvaccinated group, the bias index has been
3–8 times higher (Aaby et al., 2016).
SAGE recently reviewed the potential NSEs of BCG, MV and DTP
(Higgins et al., 2014; Strategic Advisory Group of experts on
Immunization, 2014). The reviewers indicated that the majority of stud-
ies showed a deleterious effect of DTP but they concluded that the re-
sults were inconsistent because two studies showed a beneﬁcial effect.
The beneﬁcial effect in these studies was not surprising because the
mortality rate in the unvaccinated group was unnaturally high, and
the bias index was 3.40 (2.93–3.95) and 7.52 (5.15–10.97), respectively
(Aaby et al., 2016).
SAGE's working group on non-speciﬁc effects of vaccines further
emphasized that the overall effect remains unclear because DTP has
been given in combination with other vaccines and under
circumstances where the burden of the target diseases has been re-
duced to a very low level. However, several previous studies have
shown that the negative effect of DTP-plus-OPV was not due to OPV
(Aaby et al., 2004a,b, 2012). OPV has probably reduced the overall neg-
ative effect of DTP. Previous studies have indicated that DTP (±OPV)
was associated with a 2-fold higher mortality than DTP-unvaccinated
children (Aaby et al., 2016). Since pertussis did not account for N5–6%
of infant deaths in the only existing African study of the impact of per-
tussis on child mortality (Mahieu et al., 1978), it is not surprising that
DTP is also associated with a strong negative effect prior to vaccine-in-
ducedherdimmunity(Aaby et al., 2012).
Appendix B. Supplementary Data
Supplementary data to this article can be found online at http://dx.
Aaby P, Bukh J, Lisse IM, Smits AJ, 1981. Measles vaccination and child mortality. Lancet 2: 93.
Aaby, P., Bukh, J., Lisse, I.M., Smits, A.J., 1983. Measles mortality, state of nutrition, and
family structure: a community study from Guinea-Bissau. J. Infect. Dis. 147, 693–701.
Aaby, P., Bukh, J., Lisse, I.M., Smits, A.J., 1984. Measles vaccination and reduction in child
mortality: a community study from Guinea-Bissau. J. Infect. 8, 13–21.
Aaby, P., Samb, B., Simondon, F., Knudsen, K., Coll Seck,A.M., Bennett, J., Whittle, H., 1993.
Divergent mortality for male and female recipients of low-titer and high-titer mea-
sles vaccines in rural Senegal. Am. J. Epidemiol. 138, 746–755.
Aaby, P., Samb, B., Simondon, F., Coll Seck, A.M., Knudsen, K., Whittle, H., 1995. Non-spe-
ciﬁc beneﬁcial effect of measles immunisation: analysis of mortality studies fromde-
veloping countries. Br. Med. J. 311, 481–485.
Aaby, P., Bhuyia, A., Nahar, L., Knudsen, K., de Francisco, A., Strong, M., 2003a. The survival ben-
eﬁt of measles immunization may not be explained entirely by the prevention of measles
disease: a community study from rural Bangladesh. Int. J. Epidemiol. 32, 106–115.
Aaby, P., Jensen, H., Samb, B., Cisse, B., Sodeman, M., Jakobsen, M., Poulsen, A., Rodrigues,
A., Lisse, I.M., Simondon, F., Whittle, H., 2003b. Differences in female-male mortality
after high-titre measles vaccine and association with subsequent vaccination with
diphtheria-tetanus-pertussis and inactivated poliovirus: reanalysis of West African
studies. Lancet 361, 2183–2188.
Aaby, P., Jensen, H., Gomes, J., Fernandes, M., Lisse, I.M., 2004a. The introduction of diph-
theria-tetanus-pertussis vaccine and child mortality in ruralGuinea-Bissau: an obser-
vational study. Int. J. Epidemiol. 33, 374–380.
Aaby, P., Rodrigues, A., Biai, S., Martins, C., Veirum, J.E., Benn, C.S., Jensen, H., 2004b. Oral
polio vaccination and low case fatality at the paediatric ward in Bissau, Guinea-Bis-
sau. Vaccine 22, 3014–3017.
Aaby, P., Benn, C.S., Nielsen, J., Lisse, I.M., Rodrigues, A., Ravn, H., 2012. Testing thehypoth-
esis that diphtheria-tetanus-pertussis vaccine has negative non-speciﬁcandsex-dif-
ferential effects on child survival in high-mortality countries. BMJ Open 2, e000707.
Aaby, P., Ravn, H., Benn, C.S., 2016. The WHOreview of the possible non-speciﬁc effects of
diphtheria-tetanus-pertussis vaccine. Pediatr. Infect. Dis. J. 35, 1257.
Expanded Programme on Immunization, 1982. The optimal age for measles immuniza-
tion. Wkly. Epidemiol. Rec. 57, 89–91.
Higgins, J.P.T., Soares-Weiser, K., Reingold, A., 2014. Systematic review of the non-speciﬁc
effects of BCG, DTP and measles containing va ccines. http://www.who.int/
immunization/sage/meetings/2014/april (accessed June 1, 2014).
Holt, E.A.,Boulos, R., Halsey, N.A., et al., 1990.Childhood survival in Haiti: protectiveeffect
of measles vaccination. Pediatrics 86, 188–194.
of breastfeeding after 12 months of age due to a new pregnancy and other causes is asso-
ciated with increased mortality in Guinea-Bissau. Int. J. Epidemiol. 32, 92–96.
Jensen, H.,Benn, C.S.,Lisse, I.M., Rodrigues, A., Andersen, P.K., Aaby, P., 2007. Survival bias
in observational studies of the impact of routine vaccinations on childhood survival.
Trop. Med. Int. Health 12, 5–14.
Kapoor, S.K., Reddaiah, V.P., 1991.Effectiveness of measles immunization on diarrhea and
malnutrition related mortality in 1–4 year olds. Indian J. Pediatr. 58, 821–823.
Kristensen, I., Aaby, P., Jensen, H., 2000. Routine vaccinations and childsurvival: follow up
study in Guinea-Bissau, West Africa. Br. Med. J. 321, 1435–1438.
Lund, N., Andersen, A., Hansen, A.S., Jepsen, F.S., Barbosa, A., Biering-Sørensen, S.,
Rodrigues, A., Ravn, H., Aaby, P., Benn, C.S., 2015. The effect of oral polio vaccine at
birth on infant mortality: a randomized trial. Clin. Infect. Dis. 61, 1504–1511.
Mahieu, J.M., Muller, A.S., Voorhoeve, A.M., Dikken, H., 1978. Pertussis in a rural area of Kenya:
epidemiology and a preliminary report of a vaccine trial. Bull. WHO 56, 773–780.
Strategic Advisory Group of Experts on Immunization, 2014. Wkly. Epidemiol. Rec. 89,
SAGE Non-speciﬁc Effects of Vaccines Working Group, 2014. Evidence Based Recommen-
dations on Non-speciﬁc Effects of BCG, DTP-Containing and Measles-Containing Vac-
cines on Mortality in Children under 5 years of Age. Background paper for SAGE
The Kasongo Project Team, 1981. Inﬂuence of measles vaccination on survival pattern of
7–35-month-old children in Kasongo, Zaire. Lancet i, 764–767.
198 S.W. Mogensen et al. / EBioMedicine 17 (2017) 192–198