BMJ Open Is Committed to Open Peer Review. As Part of This Commitment We Make the Peer Review History of Every Article We Publish Publicly Available

BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from

BMJ Open is committed to open peer review. As part of this commitment we make the peer review history of every article we publish publicly available.

When an article is published we post the peer reviewers’ comments and the authors’ responses online. We also post the versions of the paper that were used during peer review. These are the versions that the peer review comments apply to.

The versions of the paper that follow are the versions that were submitted during the peer review process. They are not the versions of record or the final published versions. They should not be cited or distributed as the published version of this manuscript.

BMJ Open is an open access journal and the full, final, typeset and author-corrected version of record of the manuscript is available on our site with no access controls, subscription charges or pay-per-view fees (http://bmjopen.bmj.com).

If you have any questions on BMJ Open’s open peer review process please email [email protected] http://bmjopen.bmj.com/ on September 24, 2021 by guest. Protected copyright. BMJ Open BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from

Can earlier BCG vaccination reduce early infant mortality? A cluster randomised trial

Journal: BMJ Open ManuscriptFor ID peerbmjopen-2018-025724 review only Article Type: Protocol

Date Submitted by the 30-Jul-2018 Author:

Complete List of Authors: Thysen, Sanne; Bandim Health Project, Statens Serum Institut , Jensen, Aksel; University of Copenhagen, Section of Biostatistics Rodrigues, Amabelia; Bandim Health Project, Bandim Health Project, Borges, Igualdino; Bandim Health Project, Bandim Health Project Aaby, Peter; Bandim Health Project, Benn, Christine; Statens Serum Institut, Bandim Health Project; University of Southern Denmark, OPEN Fisker, Ane; Statens Serum Institut,

BCG vaccine, Oral polio vaccine, Cluster randomised trial, Non-specific Keywords: effects of vaccines

http://bmjopen.bmj.com/ on September 24, 2021 by guest. Protected copyright.

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 1 of 31 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Can earlier BCG vaccination reduce early infant mortality? A cluster randomised trial 5 6 1,2,3 2,4 1 7 Sanne Marie Thysen , Aksel Karl Georg Jensen , Amabelia Rodrigues , Igualdino da Silva 8 1 1,2 1,2,5 1,2,5 9 Borges , Peter Aaby , Christine Stabell Benn , Ane Bærent Fisker 10 11 1Bandim Health Project, Bissau, Guinea-Bissau 12 2 13 Research Centre for Vitamins and Vaccines, Bandim Health Project, Statens Serum Institut, 14 Copenhagen, Denmark 15 16 3Center for Global Health (GloHAU), Aarhus University, Aarhus, Denmark 17 4 18 Section of Biostatistics,For Department peer of Public review Health, University only of Copenhagen, Copenhagen, 19 20 Denmark 21 5OPEN, Institute of Clinical Research, University of Southern Denmark, Odense, Denmark 22 23 24 25 Corresponding author: Sanne Marie Thysen, Bandim Health Project, INDEPTH Network, 26 27 Apartado 861, 1004 Bissau Codex, Guinea-Bissau; [email protected] 28 29 30 31 32 33 34 35 36 37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 45 on September 24, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 2 of 31

Ethics and dissemination http://bmjopen.bmj.com/ 38 39 BCG vaccination is delayed in many low-income settings. WHO-recommended home visits are 40 41 resource demanding and vaccines are not part of the recommendation. Utilizing the home visits to 42 43 provide BCG and OPV may provide countries with a further incentive to introduce a single home 44 45 visit. In countries, where home visits are already in place, vaccines can easily be added to reduce on September 24, 2021 by guest. Protected copyright. 46 early infant mortality. The trial is approved by the Guinean Ethical Committee (Reference number: 47 48 0016/CNES/INASA/2015) and the Danish Ethics Committee has given its consultative approval. 49 50 The trial is registered at clinicaltrials.gov (Clinicaltrials.gov ID: NCT02504203). The results of the 51 52 trial will be published in international peer-reviewed journals. 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 3 of 31 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Summary box 5 6 7 Strengths and limitations of the study: 8 9  The present study will provide evidence regarding the non-specific effects of BCG and OPV 10 provided at birth 11 12  The trial will be the first randomised trial to assess non-specific effects of BCG and OPV in a 13 14 rural low-income setting 15 16  The trial tests a feasible add-on intervention to the currently WHO-recommended home visits 17  The study will provide a cost-effectiveness analysis of providing BCG and OPV at a single 18 For peer review only 19 home visit after birth 20 21  The trial tests co-administering BCG and OPV at birth and will not be able to conclude on 22 23 separate non-specific effects of the individual vaccines 24 25 26 27 28 29 30 31 32 33 34 35 36 37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 45 on September 24, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 4 of 31

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 INTRODUCTION 5 6 WHO recommends home visits after birth to reduce neonatal mortality1 2. Guinea-Bissau has not 7 8 implemented such home visits yet. Vaccinations are not an integrated part of home visits and are 9 10 normally provided at the health centres. The current vaccination policy is to provide Bacillus 11 Calmette-Guérin vaccine (BCG) and oral polio vaccine (OPV) at birth (Vaccination Schedule, 12 13 Figure 1). 14 15 Public health policies in low-income countries are based on a one-disease-one-solution paradigm: 16 17 for each of the important diseases, a specific vaccine is developed. It is believed that the vaccine 18 For peer review only 19 prevents only the specific disease and does nothing else. Therefore, most routine vaccines have not 20 21 been tested for their effect on overall mortality. 22 23 However, many observational studies3-5 and randomised controlled trials (RCTs)6-9 in low-income 24 25 countries have now shown that apart from preventing specific diseases, the vaccines may also have 26 wider effects on the immune system, which leads to changes in resistance/susceptibility to unrelated 27 28 infections10. 29 30 3 31 With respect to BCG, observational studies show that BCG is associated with better child survival 32 11 12, which cannot be explained by prevention of tuberculosis (TB)13. WHO recommends BCG 33 34 vaccine at birth in low-income countries. Typically low-weight children (LW;<2500g) are only 35 36 vaccinated when they reach 2500g. These children are assumed to be immunologically immature, 37 and BCG vaccination is therefore often delayed. However, in three RCTs in Guinea-Bissau, BCG- http://bmjopen.bmj.com/ 38 39 at-birth reduced neonatal mortality in LW children, the combined analysis showed a mortality 40 41 reduction of 38% (95%CI: 17-54%)6 7. TB is very rare in early childhood14 and the effects can 42 43 therefore not be ascribed to prevention of TB. The effect of BCG-at-birth on neonatal mortality was 44 seen already during the first three days post-vaccination, the reduction in mortality being 45% (7- 45 on September 24, 2021 by guest. Protected copyright. 46 68%)9. This rapid effect suggests that the effect is due to changes in the innate immune system. 47 48 Recent immunological studies support that BCG induces epigenetic changes in monocytes, which 49 50 reprogram the innate immune system to increased pro-inflammatory responses against unrelated 51 52 pathogens15 16. These findings thus provide biological mechanisms whereby BCG could exert non- 53 16 17 54 specific beneficial effects, protecting the recipients against non-targeted infectious diseases . 55 56 Approximately 75% of neonatal deaths occur within the first week of life13, but BCG is rarely given 57 58 at birth in low-income countries; only 11% of the children receive BCG in the first week of life in 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 5 of 31 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Guinea-Bissau18. A major reason for this delay is the focus on reducing wastage of vaccines. 5 6 Freeze-dried vaccines like BCG have to be used within 6 hours after reconstitution with a diluent. 7 8 Therefore, a vial of BCG vaccine is often not opened unless 10-12 children are present for 9 10 vaccination. Hence, there are many missed vaccination opportunities. If BCG vaccine has profound 11 effects on neonatal mortality and morbidity many lives could be saved by providing BCG earlier. 12 13 14 In the present study, we aim to study the effect of BCG and OPV vaccinations provided at a single 15 home visit within 72 hours after birth on early infant mortality and morbidity among infants in a 16 17 rural and an urban setting in Guinea-Bissau. We hypothesise that BCG at birth provided at a single 18 For peer review only 19 home visit shortly after birth reduce early infant non-accidental mortality by 40%. 20 21 METHODS AND ANALYSES 22 23 Setting 24 25 The study is conducted in Bandim Health Project’s (BHP) Health and Demographic Surveillance 26 27 System (HDSS) sites in rural and urban Guinea-Bissau. 28 29 Rural site: The rural HDSS was established in 1990. The BHP teams survey women of fertile age 30 31 and their children below the age of 5 years in randomly selected clusters of villages in all the nine 32 health regions in the country. Children in the three rural regions closest to Bissau (Oio, Biombo and 33 34 Cacheu) are eligible for the study. 35 36 37 The rural clusters are followed with two-monthly visits by the mobile teams. At all visits, the http://bmjopen.bmj.com/ 38 women are asked whether they are pregnant. When a pregnancy is registered; the woman’s 39 40 nutritional status is assessed by measurement of a mid-upper-arm-circumference (MUAC); 41 42 information on antenatal care is collected prior to giving birth, as well as at the first visit after 43 delivery. Socio-economic factors (type of roofing, type of bathroom, possession of a mobile phone, 44 45 radio and generator) are registered. After the delivery, information on the place of delivery (home, on September 24, 2021 by guest. Protected copyright. 46 47 health facility) and who assisted the delivery is collected. 48 49 Urban site: The urban HDSS was established in 1978. Field assistants follow children below the 50 51 age of 3 years through home visits every third month. Pregnancies are registered every month and 52 53 followed monthly until birth. When a pregnancy is registered information on ethnic group, 54 gestational age, use of mosquito net is collected. After the delivery, information on place of delivery 55 56 (home, health facility) is collected. All children in the urban study area will be eligible for the study. 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 6 of 31

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Study nurses 5 6 Prior to study start, we selected study nurses to conduct home visits. In the rural study area, study 7 8 nurses were selected in collaboration with the regional health authorities. The nurses were selected 9 10 among nurses working at a health centre with a catchment area corresponding to the study clusters, 11 so the home visits can be done in addition to their normal routines. For the urban study area, the 12 13 BHP employed a study nurse fulltime. 14 15 Community Key Informants 16 17 In the rural areas, to ensure that the nurses are informed immediately about deliveries, we use 18 For peer review only 19 community key informants (CKI). CKIs were selected among residents in each rural village to 20 21 collect information about pregnancies, deliveries and deaths. The CKI communicate immediately 22 any delivery in the village to the study nurse. In the urban area, we take advantage of the close 23 24 follow-up by the field assistants, who circulate the zones daily. 25 26 Study design and randomisation 27 28 The trial is a cluster-randomised trial, randomising clusters (village clusters/sub-district zones) to 29 30 two different treatment groups, stratified by region and pre-trial mortality level (high/low). 31 32 Clusters, defined as village cluster in the rural area, and as sub-district zones in the urban area, were 33 34 randomised prior to study start using computer generated random numbers. All newborns are visited 35 36 as soon as possible after birth. In half of the clusters, we provide standard care and vaccines (BCG, 37 OPV) to children at the home visits; in the remaining clusters, children only receive standard care http://bmjopen.bmj.com/ 38 39 but no vaccines (Table 1). 40 41 42 Inclusion criteria 43 All children registered during pregnancy are eligible for the study. Children can only be enrolled if 44 45 they are visited within 72 hours after birth. on September 24, 2021 by guest. Protected copyright. 46 47 Exclusion criteria 48 49 There are few exclusion criteria, because the study is expected to answer a pragmatic question about 50 51 the effect of BCG and OPV vaccination at home visits shortly after birth. 52 53  Children already BCG vaccinated 54 55  Moribund children (Expected not to survive the next 24 hours, as evaluated by the study nurse 56 57 at the enrolment visit) 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 7 of 31 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4  Children in rural villages where the BHP mobile teams coincidentally were in the village the 5 6 same day (and vaccinated all children) 7 8 9 Informed consent 10 11 All pregnant women receive an oral and a written explanation of the study, and are offered to 12 participate in the study at the time of registration of the pregnancy. In the rural area, the explanation 13 14 is given by the BHP nurse at the two-monthly visits. In the urban area, the explanation is given by 15 16 the fieldwork assistant at a monthly visit. It is explained that normally newborn children receive 17 their first vaccinations at the health centres. However, studies have suggested that it might be 18 For peer review only 19 beneficial to receive BCG and OPV early and the BHP will therefore test the effect of providing the 20 21 vaccines at home visits. All children receive a home visit with standard care. Children in half the 22 23 clusters (intervention clusters) receive BCG and OPV at enrolment. Children in the control clusters 24 can seek vaccination elsewhere and will be offered the BCG vaccine and OPV at the next home 25 26 visit by the BHP team if they did not get it elsewhere. If the pregnant woman accepts to take part in 27 28 the study, they are asked to sign or fingerprint a consent form. An independent witness confirms the 29 30 consent process by co-signing the form. 31 32 Enrolment 33 34 The study nurse visits every newborn child shortly after a CKI/mother calls, if possible on the same 35 day. At the home visit, the nurse asks the mother for confirmation of her consent for participation in 36 37 the study before revealing the randomisation. If the mother is not able to present the signed consent http://bmjopen.bmj.com/ 38 39 form, new consent forms are filled in. The nurse brings a questionnaire, a sticker with study number 40 41 and a vaccination card. 42 43 Intervention 44 45 For all children, the nurse examines and weighs the child, encourages skin-to-skin contact to keep on September 24, 2021 by guest. Protected copyright. 46 the newborn warm and if necessary performs umbilical cord care1. For children in the control 47 48 clusters, the nurse informs about vaccination opportunities (vaccination at closest health centre or 49 50 vaccination by BHP nurse at next visit), as recommended by WHO1. For children in the 51 52 intervention clusters, the nurse administers BCG and OPV (Table 1). 53 54 BCG vaccine 55 56 BCG vaccination is administered by intradermal injection; after vaccination most children develop a 57 scar at the injection site. Among BCG-vaccinated children, having a BCG scar is associated with 58 59 improved survival19-21. The proportion of children developing a scar after BCG vaccination depends 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 8 of 31

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 on the vaccination technique and strain22-24. All nurses have been trained prior to the trial and 5 6 supervised intensively during the beginning of the trial to ensure correct vaccination technique. The 7 8 Tokyo 172 BCG strain from BCG Japan is used for the study. The vaccine is prequalified by WHO 9 10 and is regularly used in the national vaccination programme in Guinea-Bissau. 11 12 OPV vaccine 13 14 The OPV vaccine is supplied from the national vaccination programme and thus strain, 15 manufacturer and batch may vary. For all vaccines, the manufacturer, batch and expiration date is 16 17 registered. 18 For peer review only 19 20 Follow-up 21 All children enrolled in the study receive two follow-up visits. At follow-up visits, the child is 22 23 weighed, and the BHP nurse examines lymphadenopathy and reaction at the vaccination site. 24 25 Follow-up in the rural regions 26 27 The study participants are followed through the rural HDSS, where all children below 5 years of 28 29 age are followed, and information on vital status, breastfeeding status, supplementary feeding, 30 31 MUAC, vaccinations, hospital admissions and whether the child has received interventions 32 provided in campaigns is collected. 33 34 35 The BHP teams are accompanied by a nurse, who administers vaccines. The BHP nurse performs 36 the follow-up visits for the present study. All children have their vaccination cards examined and 37 http://bmjopen.bmj.com/ 38 those missing one or more routine vaccines according to the schedule are offered these vaccines at 39 40 the follow-up visits. 41 42 Follow-up in the urban region 43 44 The study participants receive two follow-up visits by the study nurse at 2 and 4 months of age. The 45 on September 24, 2021 by guest. Protected copyright. 46 study nurse brings BCG and OPV for children in control clusters. For the remaining routine 47 vaccines as part of the national programme, the children are referred to the closest health centre. All 48 49 children below the age of 3 years are followed through home visits every third month, where 50 51 information on vital status, breastfeeding status, supplementary feeding, MUAC, vaccinations, 52 53 hospital admissions and vaccination campaigns are collected. 54 55 Outcomes 56 57 The primary outcome is non-accidental mortality. All children living in intervention and control 58 clusters are followed through the HDSS routines with collection of information on vital status for 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 9 of 31 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 all children. Deaths are classified as accidental or non-accidental based on information from verbal 5 6 autopsies. 7 8 9 Secondary outcomes and potential effect modifiers 10 We will evaluate the effect on severe morbidity considered as the composite outcome of non- 11 12 accidental death and non-accidental hospital admission. With declining mortality, we may not be 13 14 able to obtain a conclusive result for the primary outcome. We have defined the sample size for the 15 composite outcome, but retained mortality as the primary outcome. 16 17 We will conduct secondary analyses to assess whether an effect vary by low birthweight (<2500g), 18 For peer review only 19 sex, maternal BCG scarring, and season to gain a better knowledge of the potential effect modifiers 20 21 (a list of all outcomes can be found in Table 2). 22 We will study the cost-effectiveness of providing BCG and OPV at a home visit using the effects on 23 24 mortality and hospital admission from the present project. 25 26 Data management 27 28 All data is collected on paper forms and entered in the BHP office. Records with registration of all 29 30 pregnant women are copied from the data tables of the routine HDSS to the BCG trial data table. 31 32 Forms for follow-up are pre-printed with identification information of the enrolled children. For 33 routine HDSS visits, a list of all pregnant women is printed prior to the visit to ensure that all pregnant 34 35 women are invited to participate in the study prior to giving birth. Following data entry, the data is 36 37 checked for consistency. Inconsistencies between the BCG trial data and the HDSS data are checked. http://bmjopen.bmj.com/ 38 Main outcome events are reviewed individually. 39 40 41 Statistical analyses 42 43 General analysis principles applied in all analyses can be found in appendix 1. In Cox-proportional 44 hazards models with age as underlying time-scale, we will assess the effect of providing BCG and 45 on September 24, 2021 by guest. Protected copyright. 46 OPV at a single home visit. We will allow for different baseline hazards according to the factors 47 48 used in the randomisation (region and pre-study mortality level) and sex. Children enter the analysis 49 50 on the date of enrolment or 24 hours of age, whichever comes last, and remain in the analysis until 51 the first subsequent visit by the BHP team, date of registering first non-trial vaccine after enrolment, 52 53 death or migration within the first 60 days of life. In secondary analyses, we will investigate 54 55 whether the effect of BCG+OPV differ by the following potential effect modifiers, which in prior 56 trials have been important determinants of the effect: low birthweight (<2500 g), sex, maternal 57 58 BCG scarring, and season. Further details are provided in appendix 2. 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 10 of 31

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Secondary outcomes are non-accidental hospital admissions, severe morbidity (composite outcome 5 6 of non-accidental mortality and non-accidental hospital admissions), consultations, growth (mid- 7 8 upper-arm circumference and weight-for-age z-score), and BCG scarring. Analyses are described in 9 10 appendix 2. We will furthermore perform a cost effectiveness analysis seeking to measure the cost 11 per death averted using a societal perspective, contrasting the costs of vaccine provision in the 12 13 present programme and an outreach system as tested in the trial. The final analysis plan is deposited 14 15 with the Data Safety and Monitoring Board. A list of all planned analyses is provided in appendix 2. 16 17 Sample size considerations 18 For peer review only 19 We estimated the design effect to 1.35 (ratio of square of the standard errors for the cluster- 20 21 adjusted/unadjusted HRs) based on pre-trial event data (deaths and hospitalisation) from the rural 22 HDSS clusters, where the study is conducted. We estimated the proportion of events to be 2.4% in 23 24 the absence of our intervention (prior data). Thus, in order to obtain 80% power to detect a 25 26 reduction in early infant severe morbidity if the true reduction of BCG and OPV provided at home 27 visits is larger than 40%, we will need to enrol at least 6666 children. 28 29 30 Time schedule 31 32 The trial was initiated with a pilot study in Biombo region in June 2015. The trial was scaled up to 33 34 full study including Oio and Cacheu regions in July 2016. The trial was expanded to include the 35 urban study area in July 2017. We expect enrolments to end in December 2020. 36 37 http://bmjopen.bmj.com/ 38 ETHICS AND DISSEMINATION 39 40 Ethical considerations 41 42 BCG is recommended at birth but vaccination is often delayed; in rural Guinea-Bissau most 43 44 children receive BCG after 1 month18. While there is now increasing acceptance that BCG may 45 on September 24, 2021 by guest. Protected copyright. 46 have beneficial non-specific effects, we are still a long way from altering the implementation of the 47 vaccination programme. The proposal compares two ways of delivering BCG and OPV; both are an 48 49 improvement relative to the current situation in rural Guinea-Bissau, where less than 40% of all 50 51 infants get BCG during the first month of life25. Hence, no child receives BCG later than it would 52 53 have done, had the trial not been carried out. The trial is approved by the Guinean Ethical 54 Committee (Reference number: 0016/CNES/INASA/2015) and the Danish Ethics Committee has 55 56 given its consultative approval. The trial is registered at clinicaltrials.gov (Clinicaltrials.gov ID: 57 58 NCT02504203). 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 11 of 31 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Safety monitoring 5 6 A Data Safety and Monitoring Board has been formed consisting of a paediatrician (Anja Poulsen, 7 8 Rigshospitalet, Denmark), a statistician (Morten Frydenberg, Aarhus University, Denmark) and an 9 10 epidemiologist (Torben Sigsgaard, Aarhus University, Denmark). The members have been 11 appointed on their experience, reputation for objectivity, absence of conflicts of interest, and 12 13 knowledge on clinical trial methodology. 14 15 The vaccines used for the trial are prequalified by WHO to be given at birth. Adverse reactions are 16 17 rare for both BCG and OPV. At the two first visits after enrolment in the study, the BHP nurse 18 For peer review only 19 examines the BCG vaccination site and the axillary lymph glands of all children to assess 20 21 suppurative lymphadenitis as an adverse reaction to the BCG vaccination. Other serious adverse 22 events are captured through primary and/or secondary outcomes (mortality, hospitalisations and 23 24 consultations). 25 26 Dissemination 27 28 29 The results of the study will be published in international peer-reviewed journals and results will be 30 31 communicated to the national institute of public health in Guinea-Bissau. After publication of the 32 main results on completion of the trial, data requests can be submitted to the researchers at Bandim 33 34 Health Project. 35 36 Patient and public involvement 37 http://bmjopen.bmj.com/ 38 39 The communities were involved in locating households, when the HDSS was setup and contributed 40 information allowing tracing of internal migrants between villages throughout the study period. No 41 42 participants were involved in setting the research question or the outcome measure, nor were they 43 44 involved in developing plans for recruitment, design, or implementation of the study. Local 45 on September 24, 2021 by guest. Protected copyright. 46 community key informants were selected to ensure timely information about birth. Nurses from 47 local health centres are involved in the study and have been trained to perform enrolment. This 48 49 ensures a close collaboration with the local health system. No participant was asked to advise on 50 51 interpretation or writing up the results. The results are disseminated to the national public health 52 53 institute. There are no plans to disseminate the results of the research to study participants. Study 54 results will be disseminated to local nurses involved in the study. 55 56 57 58 59 DISCUSSION 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 12 of 31

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 WHO recommends BCG and OPV at birth. However, due to the focus on not wasting vaccine 5 6 doses, BCG vaccination is often delayed18 26-28. Previous trials from Guinea-Bissau have showed 7 8 that early vaccination may have major impact on neonatal mortality among low-weight children6 7 9, 9 10 and observational studies have suggested that BCG also has beneficial NSEs among normal-birth- 11 weight children3 11. In 2014, the Strategic Advisory Group of Experts on Immunization reviewed 12 13 the non-specific effects of the BCG vaccine, and concluded that the evidence was consistent with 14 15 beneficial NSEs. However, the bulk of evidence came from observational studies and they therefore 16 called for randomised trials. To our knowledge, this will be the first randomised trial assessing the 17 18 NSEs of BCG vaccinationFor among peer children inreview a rural low-income only setting. The SAGE review did not 19 20 include OPV, but a recent trial from Guinea-Bissau randomised children to BCG only or BCG and 21 29 22 OPV, and found that OPV was associated with a 32% reduction in mortality . Thus, both BCG and 23 OPV may have beneficial NSEs. The results will therefore be very important in broadening our 24 25 understanding of the NSEs of BCG and OPV at birth. 26 27 WHO recommends three home visits after birth, which is resource demanding and not implemented 28 29 in many low-income countries. It is not recommended that the home visits are utilised for early 30 31 vaccination. However, we hypothesise that a single home visit after birth will have major effects on 32 33 early infant mortality if BCG and OPV vaccination are added to the home visit. If BCG and OPV 34 provided at a single home visit can reduce early infant mortality, this should be included in the 35 36 WHO recommendations. A single home visit with vaccines will be easier to implement in a strained 37 http://bmjopen.bmj.com/ 38 health system than three home visits within 7 days. In countries, where home visits are already in 39 40 place, vaccines can easily be added to reduce early infant mortality. 41 42 43 44 45 on September 24, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 13 of 31 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 References 5 6 7 1. WHO, UNICEF. WHO/Unicef Joint Statement: Home visits for the newborn child: a strategy to improve 8 survival, 2009. 9 2. WHO. WHO recommendations on postnatal care of the mother and newborn. 10 http://apps.who.int/iris/bitstream/handle/10665/97603/9789241506649_eng.pdf?sequence=1, 11 2013. 12 3. Kristensen I, Aaby P, Jensen H. Routine vaccinations and child survival: follow up study in Guinea-Bissau, 13 West Africa. BMJ 2000;321(7274):1435-8. [published Online First: 2000/12/09] 14 15 4. Aaby P, Benn C, Nielsen J, et al. Testing the hypothesis that diphtheria-tetanus-pertussis vaccine has 16 negative non-specific and sex-differential effects on child survival in high-mortality countries. BMJ 17 open 2012;2(3) doi: 10.1136/bmjopen-2011-000707 18 5. Aaby P, Ravn H, Roth A,For et al. Early peer diphtheria-tetanus-pertussis review onlyvaccination associated with higher 19 female mortality and no difference in male mortality in a cohort of low birthweight children: an 20 observational study within a randomised trial. Archives of Disease in Childhood 2012;97(8):685-91. 21 doi: 10.1136/archdischild-2011-300646 22 6. Aaby P, Roth A, Ravn H, et al. Randomized trial of BCG vaccination at birth to low-birth-weight children: 23 24 beneficial nonspecific effects in the neonatal period? J Infect Dis 2011;204(2):245-52. doi: 25 10.1093/infdis/jir240 [published Online First: 2011/06/16] 26 7. Biering-Sorensen S, Aaby P, Napirna BM, et al. Small randomized trial among low-birth-weight children 27 receiving bacillus Calmette-Guerin vaccination at first health center contact. Pediatr Infect Dis J 28 2012;31(3):306-8. doi: 10.1097/INF.0b013e3182458289 29 8. Aaby P, Jensen H, Samb B, et al. Differences in female-male mortality after high-titre measles vaccine and 30 association with subsequent vaccination with diphtheria-tetanus-pertussis and inactivated 31 poliovirus: reanalysis of West African studies. Lancet 2003;361(9376):2183-8. doi: 10.1016/S0140- 32 6736(03)13771-3 [published Online First: 2003/07/05] 33 34 9. Biering-Sørensen S, Aaby P, Lund N, et al. Early BCG-Denmark and Neonatal Mortality Among Infants 35 Weighing <2500 g: A Randomized Controlled Trial. Clinical Infectious Diseases 2017;65(7):1183-90. 36 doi: 10.1093/cid/cix525 37 10. Benn CS, Netea MG, Selin LK, et al. A small jab - a big effect: nonspecific immunomodulation by http://bmjopen.bmj.com/ 38 vaccines. Trends in immunology 2013;34(9):431-9. doi: 10.1016/j.it.2013.04.004 39 11. Hirve S, Bavdekar A, Juvekar S, et al. Non-specific and sex-differential effects of vaccinations on child 40 survival in rural western India. Vaccine 2012;30(50):7300-8. doi: 10.1016/j.vaccine.2012.09.035 41 [published Online First: 2012/10/02] 42 43 12. Aaby P, Kollmann TR, Benn CS. Nonspecific effects of neonatal and infant vaccination: public-health, 44 immunological and conceptual challenges. Nature immunology 2014;15(10):895-9. doi: 45 10.1038/ni.2961 [published Online First: 2014/09/19] on September 24, 2021 by guest. Protected copyright. 46 13. Lawn JE, Cousens S, Zupan J. "4 million neonatal deaths: When? Where? Why?". Lancet 2005;365:891- 47 900. 48 14. Shann F. Commentary: BCG vaccination halves neonatal mortality. Pediatr Infect Dis J 2012;31(3):308-9. 49 doi: 10.1097/INF.0b013e3182465be8 [published Online First: 2012/02/15] 50 15. Netea MG, Quintin J, van der Meer JW. Trained immunity: a memory for innate host defense. Cell host 51 52 & microbe 2011;9(5):355-61. doi: 10.1016/j.chom.2011.04.006 [published Online First: 2011/05/18] 53 16. Kleinnijenhuis J, Quintin J, Preijers F, et al. Bacille Calmette-Guerin induces NOD2-dependent 54 nonspecific protection from reinfection via epigenetic reprogramming of monocytes. Proceedings 55 of the National Academy of Sciences of the United States of America 2012;109(43):17537-42. doi: 56 10.1073/pnas.1202870109 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 14 of 31

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 17. Jensen KJ, Larsen N, Biering-Sorensen S, et al. Heterologous immunological effects of early BCG 5 vaccination in low-birth-weight infants in Guinea-Bissau: a randomized-controlled trial. J Infect Dis 6 7 2015;211(6):956-67. doi: 10.1093/infdis/jiu508 [published Online First: 2014/09/12] 8 18. Thysen SM, Byberg S, Pedersen M, et al. BCG coverage and barriers to BCG vaccination in Guinea- 9 Bissau: an observational study. BMC Public Health 2014;14:1037. doi: 10.1186/1471-2458-14-1037 10 19. Roth A, Gustafson P, Nhaga A, et al. BCG vaccination scar associated with better childhood survival in 11 Guinea-Bissau. Int J Epidemiol 2005;34(3):540-7. doi: 10.1093/ije/dyh392 [published Online First: 12 2005/01/22] 13 20. Roth A, Sodemann M, Jensen H, et al. Tuberculin reaction, BCG scar, and lower female mortality. 14 Epidemiology 2006;17(5):562-8. doi: 10.1097/01.ede.0000231546.14749.ab [published Online First: 15 16 2006/08/01] 17 21. Garly ML, Martins CL, Bale C, et al. BCG scar and positive tuberculin reaction associated with reduced 18 child mortality inFor West Africa. peer A non-specific review beneficial effect only of BCG? Vaccine 2003;21(21- 19 22):2782-90. doi: S0264410X03001816 [pii] [published Online First: 2003/06/12] 20 22. Roth A, Sodemann M, Jensen H, et al. Vaccination technique, PPD reaction and BCG scarring in a cohort 21 of children born in Guinea-Bissau 2000-2002. Vaccine 2005;23(30):3991-8. doi: 22 10.1016/j.vaccine.2004.10.022 [published Online First: 2005/05/19] 23 23. Frankel H, Byberg S, Bjerregaard-Andersen M, et al. Different effects of BCG strains - A natural 24 experiment evaluating the impact of the Danish and the Russian BCG strains on morbidity and scar 25 26 formation in Guinea-Bissau. Vaccine 2016;34(38):4586-93. doi: 10.1016/j.vaccine.2016.07.022 27 24. Funch KM, Thysen SM, Rodrigues A, et al. Determinants of BCG scarification among children in rural 28 Guinea-Bissau: A prospective cohort study. Human vaccines & immunotherapeutics 2018:1-9. doi: 29 10.1080/21645515.2017.1421879 [published Online First: 2018/01/03] 30 25. Thysen SM, Byberg S, Pedersen M, et al. BCG coverage and barriers to BCG vaccination in Guinea- 31 Bissau: an observational study. BMC Public Health 2014;14(1):1037. doi: 10.1186/1471-2458-14- 32 1037 [published Online First: 05-10-2014] 33 26. Kagone M, Ye M, Nebie E, et al. Vaccination coverage and factors associated with adherence to the 34 35 vaccination schedule in young children of a rural area in Burkina Faso. Global health action 36 2017;10(1):1399749. doi: 10.1080/16549716.2017.1399749 [published Online First: 2017/12/01] 37 27. Hanifi SMA, Das S, Rahman M. Bangladeshi neonates miss the potential benefits of early BCG http://bmjopen.bmj.com/ 38 vaccination. Int J Epidemiol 2018;47(1):348-49. doi: 10.1093/ije/dyx223 [published Online First: 39 2017/11/01] 40 28. Wallace AS, Willis F, Nwaze E, et al. Vaccine wastage in Nigeria: An assessment of wastage rates and 41 related vaccinator knowledge, attitudes and practices. Vaccine 2017;35(48 Pt B):6751-58. doi: 42 10.1016/j.vaccine.2017.09.082 [published Online First: 2017/10/27] 43 44 29. Lund N, Andersen A, Hansen AS, et al. The Effect of Oral Polio Vaccine at Birth on Infant Mortality: A 45 Randomized Trial. Clinical infectious diseases : an official publication of the Infectious Diseases on September 24, 2021 by guest. Protected copyright. 46 Society of America 2015;61(10):1504-11. doi: 10.1093/cid/civ617 [published Online First: 47 2015/07/30] 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 15 of 31 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Authors’ contributions 5 6 AR, PA and CB conceived the idea for the trial. SMT, PA, CB and ABF designed the study. SMT 7 8 set up the study with help from ABF. SMT, IDSB and ABF supervised data collection. SMT, 9 10 AKGJ, and ABF planned the data analyses. SMT drafted the manuscript and data analysis plan with 11 help from AKGJ and ABF. All authors read and approved the final manuscript and data analysis 12 13 plan. 14 15 16 Funding 17 Karen Elise Jensen foundation is the main funder of the trial. Laerdals Foundation supports the trial. 18 For peer review only 19 The Research Centre for Vitamins and Vaccines, Bandim Health Project receives support from the 20 21 Danish National Research Foundation [DNRF108]. The funders had no role in the design of the 22 23 study or writing of the protocol. 24 25 Competing interests 26 27 The authors declare no competing interests. 28 29 30 31 32 33 34 35 36 37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 45 on September 24, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from

BMJ Open Page 16 of 31

1 2 3 4 Table 1: Trial design 5 6 Routine visit 2nd and 3rd routine visit 7 Timing before birth Shortly after birth – within 72 hours 1st routine visit after birth after birth 8 9 BHP Field Community Key 10 Personnel assistant and Nurse Informant (CKI) Study Nurse BHP Field assistant and Nurse BHP Field assistant and Nurse 11 12 Control group Information about CKI inform Field assistant: trial and invitationFor about peer birth  Umbilicalreview cord only 13  Interview on morbidity and 14 to participate. and skin care 15 Informed consent.  Encourage skin- mortality 16 to-skin contact 17 to keep the Nurse: http://bmjopen.bmj.com/ 18 newborn warm  Examine the BCG 19  Examine and vaccination site and lymph 20 glands 21 Weigh the child 22  Inform about  Measure temperature 23 vaccination  Weigh the child Field assistant: 24 opportunities  Provide BCG and OPV 25 Intervention Field assistant: on September 24, 2021 by guest. Protected copyright.  Interview on morbidity and 26 group  Umbilical cord mortality 27 and skin care  Interview on morbidity and 28  Encourage skin- mortality Nurse: 29  Examine the BCG 30 to-skin contact 31 to keep the Nurse: vaccination site and lymph 32 newborn warm  Examine the BCG glands 33  Examine and vaccination site and lymph  Measure temperature 34 weigh the child glands  Weigh the child 35  Provide BCG  Measure temperature  Provide routine 36 37 and OPV  Weigh the child vaccinations 38 39 40 41 42 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 17 of 31 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Table 2: Outcomes 5 6 Primary outcome 7 8  Early infant non-accidental mortality 9 Potential effect modifiers 10  Low birthweight defined as birthweight lower than 11 2500g 12  Sex 13 14  Maternal BCG scarring 15  Season 16 Secondary outcomes 17  Non-accidental hospital admission 18 For peer review only 19  Severe morbidity (composite outcome of non-accidental 20 mortality and non-accidental hospital admissions) 21  Consultations 22  Growth 23  Mid-upper-arm circumference 24 25  Weight-for-age z-score 26  BCG scarring 27  Cost-effectiveness of providing BCG and OPV at home 28 visits 29 30 31 32 33 34 35 36 37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 45 on September 24, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 18 of 31 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For peer review only 17 18 19 20 21 22 Figure 1: Recommended vaccination schedule in Guinea-Bissau 23 24 25 26 27 28 29 30 31 32

33 http://bmjopen.bmj.com/ 34 35 36 37 38 39 40

41 on September 24, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 19 of 31 BMJ Open

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 5 6 Appendix 1: General analysis principles 7 8 1 Participant population 9 All main analyses will be completed in a per-protocol (PP) population of all BCG-unvaccinated 10 11 children, who received a home visit within 72 hours after birth, and who received the assigned 12 13 treatment. The primary analysis and hence the main conclusion of the trial will be based on the PP 14 analysis. Unless explicitly stated all analyses will be PP. 15 16 17 Since the cluster-size varies, data will be analysed on individual level data. All statistical tests will 18 For peer review only 19 be 2-tailed and p<=0.05 considered statistically significant for analyses involving the primary 20 outcome. 21 22 23 2 Unadjusted and adjusted analyses 24 Both unadjusted analyses and analyses adjusted for place of delivery will be reported. Conclusions 25 26 will be based on the unadjusted analyses. 27 28 29 3 Multiple testing 30 P-values will not be corrected for multiple testing. Secondary outcomes are tested to observe if the 31 32 pattern is similar across other health outcomes. Consequently, p<=0.05 will not be employed as a 33 34 threshold for statistical significance for secondary outcomes. For the sensitivity analyses, we will 35 not consider statistical significance, but rather robustness of the conclusions across different 36 37 definitions of outcomes and co-variates. http://bmjopen.bmj.com/ 38 39 40 4 Missing data 41 All analyses will be complete-case analyses. 42 43 44 5 Proportional hazards 45 To test the proportional hazards assumption, a required assumption of the Cox regression, we will on September 24, 2021 by guest. Protected copyright. 46 47 perform formal significance tests based on Schoenfeld residuals. In addition, we will assess 48 49 proportionality by allowing the hazard ratio to interact with the underlying timescale to identify a 50 51 possible time trend. Finally, we will assess proportionality graphically via log-log survival curves. 52 53 Significance tests based on Schoenfeld residuals will be performed via the stata command estat 54 55 phtest, detail leading to both a global test and a test for each covariate, the latter being relevant only 56 when we study effect modifications. Presentation of log-log survival curves will be undertaken via 57 58 stphplot. Finally, possible interactions between hazard ratios and the underlying time scale will be 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 20 of 31

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 further investigated via the stcox procedure and the tvc() option. For the models including effect 5 6 modifications we will construct a new interaction variable (i.e., a four-level variable representing 7 8 the interaction) such that a graphical assessment of proportionality can be undertaken assessing the 9 10 four-level variable in a log-log survival plot. 11 12 If we identify evidence for non-proportionality, we will still report the marginal hazard ratios but 13 14 supplement this measure by hazard ratios for 2-3 properly selected categorical time-periods 15 identified based on the aforementioned proportionality investigations. 16 17 18 For peer review only 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 45 on September 24, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 21 of 31 BMJ Open

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 5 6 Appendix 2 - Planned analyses 7 8 1 Baseline 9 Descriptive statistics: 10 11 For eligible children visited by a study nurse within 72 hours after birth, we will describe reasons 12 13 for exclusion by group allocation. 14 15 Distribution of background factors will be presented by group allocation overall and by sex and 16 17 region. Background factors will be summarised by counts (percentages), means (standard deviation) 18 For peer review only 19 or medians (interquartile range) as appropriate. Information on the proportion with missing 20 information will be provided. 21 22 23 Table 1: Summary of background factors by intervention and control group 24 • 25 Sex 26 • Age at enrolment 27 • Region 28 29 • Weight at enrolment 30 • Temperature 31 • 32 Mid-upper-arm circumference 33 • Head circumference 34 • Place of birth 35 36 • Socioeconomic factors (maternal education and housing conditions) 37 http://bmjopen.bmj.com/ 38 39 40 2 Primary analysis of primary outcome 41 42 The primary analysis of early infant non-accidental mortality will be assessed on a per-protocol 43 44 (PP) analysis allowing for different baseline hazards according to factors used in the randomization 45 (Region, pre-study mortality level (high/low)) and sex, thus allowing different baseline hazards for on September 24, 2021 by guest. Protected copyright. 46 47 boys and girls. To account for clustering we will employ cluster-robust variance estimates. 48 49 For the primary outcome, we will use Cox proportional hazards models that allow different baseline 50 hazards according to the above mentioned factors and with age as underlying time-scale. Deaths 51 52 due to accidents will be censored. 53 54 The primary analysis of the primary outcome is described in more detail in table 2. 55 56 Table 2: Primary analysis of primary outcome 57 Population Children visited within 72 hours after birth are eligible for the study. 58 59 Exclusion criteria: 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 22 of 31

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 - Children already BCG vaccinated 5 6 - Moribund children (Expected not to survive the next 24 hours, 7 as evaluated by the health facility nurse at the enrolment visit) 8 - Children in rural villages where the BHP mobile teams 9 10 coincidentally were in the village the same day (and 11 vaccinated all children) 12 Observation period From: Enrolment visit or 24 hours after birth, whichever comes last 13 14 To: 60 days of life 15 Censoring, first of: 16 - Visit by the BHP 17

18 For peer- 60 days review only 19 - Death due to accident 20 - Date of registering first non-trial vaccine given after 21 22 enrolment 23 - Migration (migration out of the study area as per HDSS 24 definition) 25 26 Time scale Age 27 Failure definition Death 28 29 Statistical tool Cox proportional hazards model 30 Stratification We will employ models that allow for different baseline hazards 31 according to sex, region and pre-study mortality level 32 33 Clustering We will use robust standard errors to account for clustering 34 Outline stata code 35 For analysis: stset outdate, f(dead==1) enter (max(dob+1, date_enrol) /// 36 37 exit (min ( dnasc+60, date_mobileteam) origin(dob) http://bmjopen.bmj.com/ 38 39 stcox random, strata(sex reg prmorlev) vce(cl regam) 40 41 42 For model checking: estat phtest, detail† 43 stphplot, strata(random) adj(sex reg prmorlev) 44 45 stcox random, strata(sex reg prmorlev) vce(cl regam) tvc(random) on September 24, 2021 by guest. Protected copyright. 46 texp(_t) 47 48 *reg=region, prmorlev=pre-study mortality level, regam=cluster 49 50 † If the proportional hazards assumption is not fulfilled we will still report the marginal hazard ratio 51 but it will be supplemented by hazard ratios for two-three categorized time periods. 52 53 54 3 Effect-modifier analyses of primary outcome 55 56 We will assess whether the effect of the intervention on the primary effect measure is modified by 57 the following potential effect modifiers. 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 23 of 31 BMJ Open

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Table 3. Sex as a potential effect modifier of the primary outcome 5 6 Potential effect Sex 7 modifier 8 9 Design We will perform the analysis as describe above (for the primary 10 analysis) allowing the effect of the intervention to differ between the 11 sexes. 12 1 2 13 Reasoning Previous studies have found sex-differential non-specific effects , 14 therefore, we will assess the sex-differential effects. 15 16 Outline stata code: 17 For analysis: stcox random#sex sex, strata(sex reg prmorlev) vce(cl regam) 18 Forcontrast peer random#sex review only 19 20

21 For model checking: estat phtest, detail† 22 stphplot, strata(random_sex) adj(sex reg prmorlev) 23 stcox random#sex sex, strata(sex reg prmorlev) vce(cl regam) /// 24 25 tvc(random#sex sex) texp(_t) 26 27 *reg=region, prmorlev=pre-study mortality level, regam=cluster, 28 29 random_sex=a four-level variable based on the four possible 30 combinations of random and sex 31 † If the proportional hazards assumption is not fulfilled we will still report the marginal hazard ratio 32 33 but it will be supplemented by hazard ratios for two-three categorized time periods. 34 35 Table 4. Maternal BCG scar as a potential effect modifier of the primary outcome 36 37 Potential effect Maternal BCG scar (yes/no) http://bmjopen.bmj.com/ 38 39 modifier 40 Design We will perform the analysis as described above (for the primary 41 analysis) allowing the effect of the intervention to differ by maternal 42 43 BCG-scar status 44 Reasoning A recent randomized trial in Denmark found that the effect of BCG 45 3 varied by whether the mother had received BCG or not . Since, BCG on September 24, 2021 by guest. Protected copyright. 46 47 scar is a life-long marker of a successful BCG-vaccination, we will 48 assess whether the effect of BCG differs by maternal BCG-scar status 49 Outline stata code: 50 51 For analysis: stcox random#mBCGscar mBCGscar, strata(sex reg prmorlev) vce(cl 52 regam) 53 contrast random#mBCGscar 54 55 For model checking:

56 estat phtest, detail† 57 stphplot, strata(random_mBCGscar) adj(sex reg prmorlev) 58 stcox random#mBCGscar mBCGscar, strata(sex reg prmorlev) /// 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 24 of 31

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 vce(cl regam) tvc(random#mBCGscar mBCGscar) texp(_t) 5 6 7 * mBCGscar= maternal BCG scar, reg=region, prmorlev=pre-study 8 mortality level, regam=cluster, random_mBCGscar=a four-level 9 10 variable based on the four possible combinations of random and 11 maternal BCG scar 12 † If the proportional hazards assumption is not fulfilled we will still report the marginal hazard ratio 13 14 but it will be supplemented by hazard ratios for two-three categorized time periods. 15 16 Table 5. Low birthweight as a potential effect modifier of the primary outcome 17 Potential effect Low birthweight (<2500g: yes/no) 18 For peer review only 19 modifier 20 Design We will perform the analysis as described above (for the primary 21 analysis) allowing the effect of the intervention to differ by birthweight 22 23 strata 24 Reasoning Birthweight is an important risk factor for mortality. Previous 25 randomised trials from Guinea-Bissau assessing the effect of BCG on 26 27 mortality have been performed among low-birth-weight children. 28 Outline stata code: 29 For analysis: stcox random#LBW LBW, strata(sex reg prmorlev) vce(cl regam) 30 31 contrast random#LBW 32 33 For model checking: estat phtest, detail† 34 35 stphplot, strata(random_LBW) adj(sex reg prmorlev) 36 stcox random#LBW LBW, strata(sex reg prmorlev) vce(cl regam) /// 37 tvc(random#LBW LBW) texp(_t) http://bmjopen.bmj.com/ 38 39 40 * LBW=Low birthweight, reg=region, prmorlev=pre-study mortality 41 level, regam=cluster, random_LBW=a four-level variable based on the 42 43 four possible combinations of random and low birthweight 44 † If the proportional hazards assumption is not fulfilled we will still report the marginal hazard ratio 45 but it will be supplemented by hazard ratios for two-three categorized time periods. on September 24, 2021 by guest. Protected copyright. 46 47 48 Table 6. Season as a potential effect modifier of the primary outcome 49 50 Potential effect Season of birth (Dry: December-May/Rainy: June-November) 51 modifier 52 53 Design We will perform the analysis as described above (for the primary 54 analysis) allowing the effect of the intervention to differ by season of 55 birth 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 25 of 31 BMJ Open

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Reasoning Previous studies have found that the effect of some vaccines is stronger 5 4 6 in the dry season . Therefore, we would like to assess if the effect of 7 BCG and OPV differs according to season. 8 Outline stata code: 9 10 For analysis: stcox random#season season, strata(sex reg prmorlev) vce(cl regam) 11 contrast random#season 12 13

14 For model checking: estat phtest, detail† 15 stphplot, strata(random_season) adj(sex reg prmorlev) 16 stcox random#season season, strata(sex reg prmorlev) vce(cl regam) /// 17 18 Fortvc(random#season peer review season) texp(_t) only 19 20 *reg=region, prmorlev=pre-study mortality level, regam=cluster, 21 22 random_season=a four-level variable based on the four possible 23 combinations of random and season 24 † If the proportional hazards assumption is not fulfilled we will still report the marginal hazard ratio 25 but it will be supplemented by hazard ratios for two-three categorized time periods. 26 27 28 29 4 Primary analyses of secondary outcomes 30 Non-accident morbidity 31 32 Since it can be difficult to distinguish between hospital admissions and outpatient contact through 33 34 interviews, we have defined hospital admissions as an overnight stay in a health facility. Hospital 35 36 admissions due to accidents are ignored but the follow-up time is (interval) censored while the child 37 is admitted. http://bmjopen.bmj.com/ 38 39 Table 7: Non-accident hospitalisation 40 41 Population Identical to primary analysis of primary outcome 42 Observation period From: Enrolment visit or 24 hours after birth, whichever comes last 43 To: 60 days of life 44 45 Censoring, first of: on September 24, 2021 by guest. Protected copyright. 46 - Visit by the BHP 47 - Date of registering first non-trial vaccine given after 48 49 enrolment 50 - Death 51 - Hospital admission due to accident 52 53 - 60 days 54 - Migration (migration out of the study area as per HDSS 55 definition) 56 57 Time scale Age 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 26 of 31

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Failure definition First hospital admission – only overnight hospitalisations or arrival at 5 6 the hospital and death within the first day will be considered in this 7 analysis. 8 Statistical tool Cox proportional hazards model 9 10 Stratification We will employ models that allow for different baseline hazards 11 according to sex, region and pre-study mortality level 12 Clustering We will use robust standard errors to account for clustering 13 14 Outline stata code 15 For analysis: stset outdate, f(hosp==1) enter (max(dob+1, date_enrol) /// 16 exit (min ( dnasc+60, date_mobileteam) origin(dob) 17 18 For peer review only 19 stcox random, strata(sex reg prmorlev) vce(cl regam) 20 21

22 For model checking: estat phtest, detail† 23 stphplot, strata(random) adj(sex reg prmorlev) 24 stcox random, strata(sex reg prmorlev) vce(cl regam) tvc(random) 25 26 texp(_t) 27 28 *reg=region, prmorlev=pre-study mortality level, regam=cluster 29 30 † If the proportional hazards assumption is not fulfilled we will still report the marginal hazard ratio 31 but it will be supplemented by hazard ratios for two-three categorized time periods. 32 33 Severe morbidity 34 35 We will evaluate the effect on severe morbidity considered as the composite outcome of non- 36 accidental death and non-accidental hospital admission. Since it can be difficult to distinguish 37 http://bmjopen.bmj.com/ 38 between hospital admissions and outpatient contact through interviews, we have defined hospital 39 40 admissions as an overnight stay in a health facility. Hospital admissions due to accidents are 41 42 ignored but the follow-up time is (interval) censored while the child is admitted. The potential effect 43 modifiers for the primary outcome specified in tables 3-6 will also be assessed for the composite 44 45

outcome. on September 24, 2021 by guest. Protected copyright. 46 47 Table 8: Severe morbidity 48 49 Population Identical to primary analysis of primary outcome 50 Observation period From: Enrolment visit or 24 hours after birth, whichever comes last 51 To: 60 days of life 52 53 Censoring, first of: 54 - Visit by the BHP 55 - Date of registering first non-trial vaccine given after 56 57 enrolment 58 - Death/Hospital admission due to accident 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 27 of 31 BMJ Open

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 - 60 days 5 6 - Migration (migration out of the study area as per HDSS 7 definition) 8 Time scale Age 9 10 Failure definition Death or first hospital admission 11 12 Statistical tool Cox proportional hazards model 13 Stratification We will employ models that allow for different baseline hazards 14 according to sex, region and pre-study mortality level 15 16 Clustering We will use robust standard errors to account for clustering 17 Outline stata code: 18 For analysis Forstset peer outdate, f(event==1) review enter (max(dob+1,only date_enrol) /// 19 20 exit (min ( dnasc+60, date_mobileteam) origin(dob) 21 22 stcox random, strata(sex reg prmorlev) vce(cl regam) 23 24

25 For model checking estat phtest, detail† 26 stphplot, strata(random) adj(sex reg prmorlev) 27 28 stcox random, strata(sex reg prmorlev) vce(cl regam) tvc(random) 29 texp(_t) 30 31 *reg=region, prmorlev=pre-study mortality level, regam=cluster 32 33 † If the proportional hazards assumption is not fulfilled we will still report the marginal hazard ratio 34 but it will be supplemented by hazard ratios for two-three categorized time periods. 35 36 37 Table 9: All-cause consultations http://bmjopen.bmj.com/ 38 39 Population Identical to primary analysis of primary outcome 40 Observation period From: Enrolment visit or 24 hours after birth, whichever comes last 41 To: 60 days of life 42 43 Censoring, first of: 44 - Visit by the BHP 45 - Date of registering first non-trial vaccine given after on September 24, 2021 by guest. Protected copyright. 46 47 enrolment 48 - 60 days 49 - Migration (migration out of the study area as per HDSS 50 51 definition) 52 - Death 53 - End of study 54 55 Failure definition An out-patient consultation within the observation period 56 Statistical tool Log-binomial regression 57 Adjustment We will adjust the analysis for sex, region and pre-study mortality 58 59 level 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 28 of 31

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Clustering We will use robust standard errors to account for clustering 5 6 Stata code Binreg cons random sex b1.reg prmorlev, rr vce(cl regam) 7 *cons=out-patient consultation, reg=Region, prmorlev=Pre-study 8 mortality level, regam=cluster 9 10 11 12 Growth 13 Table 10: Mid-upper-arm circumference (MUAC) 14 15 Population Identical to primary analysis of primary outcome 16 17 Observation time point First visit by the mobile teams 18 Growth measures For MUAC peer will be review analysed using the only measured value 19 Statistical tool Linear regression 20 21 Adjustment We will adjust the analysis for MUAC at enrolment, age at MUAC 22 measurement, sex, region and pre-study mortality level 23 Clustering We will use robust standard errors to account for clustering 24 25 Stata code Regress MUAC random sex b1.reg prmorlev MUACenrol 26 MUACage, /// vce(cl regam) 27 *reg=region, prmorlev=pre-study mortality level, 28 29 MUACenrol=MUAC at enrolment, MUACage= age at MUAC 30 assessment, regam=cluster 31 32 33 Table 11: Weight-for-age z-score 34 35 Population Identical to primary analysis of primary outcome 36 Observation time point First visit by the mobile teams 37 http://bmjopen.bmj.com/ 38 Growth measures Weight will be analysed using the WHO weight-for-age z-score 39 Statistical tool Linear regression 40 Adjustment We will adjust the analysis for weight-for-age at enrolment, sex, 41 42 region and pre-study mortality level 43 Clustering We will use robust standard errors to account for clustering 44 Stata code Regress w-z-score random w-z-enrol sex b1.reg prmorlev, vce(cl 45 on September 24, 2021 by guest. Protected copyright. 46 regam) 47 *w-z-score=weight-for-age z-score at first visit by the mobile teams, 48 w-z-enrol=weight-for-age z-score at enrolment, reg=region, 49 50 prmorlev=pre-study mortality level, regam=cluster 51 52 53 BCG scarring 54 55 Table 12: BCG scarring 56 57 Population Identical to primary analysis of primary outcome 58 Observation timepoint First visit after 6 months of age 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 29 of 31 BMJ Open

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Failure definition Scar (yes/no) 5 6 Statistical tool Log-binominal regression 7 Adjustment We will adjust the analysis for sex, region and pre-study mortality 8 level 9 10 Clustering We will use robust standard errors to account for clustering 11 Stata code Binreg scar random sex b1.reg prmorlec, rr vce(cl regam) 12 13 *reg=region, prmorlev=pre-study mortality level, regam=cluster 14 15 16 Cost-effectiveness of providing BCG and OPV at birth 17 18 A cost effectiveness analysisFor seeking peer to measure review the cost per deathonly averted using a societal 19 perspective will be performed, contrasting the costs of vaccine provision in the present programme 20 21 and an outreach system as tested in the trial. The costs/savings associated with different rates of 22 23 consultations and admissions will also be taken into account. 24 25 Suppurative lymphadenitis 26 27 We will assess the incidence of suppurative lymphadenitis as a reaction to BCG vaccination in the 28 29 intervention and control clusters. Other serious adverse events to the BCG and OPV vaccine will be 30 captured through the outcome measures (mortality, hospital admission and consultations). 31 32 33 5 Sensitivity analyses to test for robustness of conclusions 34 35 Table 13: Cause-specific death 36 Population Identical to primary analysis of primary outcome 37 http://bmjopen.bmj.com/ 38 Observation period From: Enrolment visit or 24 hours after birth, whichever comes last 39 To: 60 days of life 40 Censoring, first of: 41 42 - Visit by the BHP 43 - 60 days 44 - Death due to accident 45 on September 24, 2021 by guest. Protected copyright. 46 - Date of registering first non-trial vaccine given after 47 enrolment 48 - Migration (migration out of the study area as per HDSS 49 50 definition) 51 Time scale Age 52 Failure definition Death due to: 53 54 Malaria, Respiratory Infection, Sepsis, Gastrointestinal disease, Other 55 Statistical tool Cox proportional hazards model 56 Stratification We will employ models that allow for different baseline hazards 57 58 according to sex, region and pre-study mortality level 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 30 of 31

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Clustering We will use robust standard errors to account for clustering 5 6 Outline stata code 7 For analysis: stset outdate, f(event==1&cause==X) enter (max(dob+1, date_enrol) 8 /// exit (min (dnasc+60, date_mobileteam) origin(dob) 9 10 11 stcox random, strata(sex reg prmorlev) vce(cl regam) 12 13

14 For model checking: estat phtest, detail† 15 stphplot, strata(random) adj(sex reg prmorlev) 16 stcox random, strata(sex reg prmorlev) vce(cl regam) tvc(random) 17 18 Fortexp(_t) peer review only 19 20 *reg=region, prmorlev=pre-study mortality level, regam=cluster 21 22 † If the proportional hazards assumption is not fulfilled we will still report the marginal hazard ratio 23 but it will be supplemented by hazard ratios for two-three categorized time periods. 24 25 26 In sensitivity analyses, we will furthermore, assess whether the conclusions are robust to the 27 28 following: 29 30 - Censoring follow-up at general health intervention campaigns (e.g. OPV campaigns) 31 32 - Altering the population to using an intention-to-treat approach, including all children, who 33 had a home visit by a study nurse (i.e., including children who did not receive assigned 34 35 treatment, children who were not enrolled because they were moribund, or who did not 36 37 accept to participate). http://bmjopen.bmj.com/ 38 39 - For the non-accident hospitalisations, we will perform the analyses allowing for repeated 40 hospitalisations. A child will return to the at-risk population the day after discharge from the 41 42 hospital. 43 44 45 References on September 24, 2021 by guest. Protected copyright. 46 47 1. Benn CS, Netea MG, Selin LK, et al. A small jab - a big effect: nonspecific immunomodulation by vaccines. 48 Trends in immunology 2013;34(9):431-9. doi: 10.1016/j.it.2013.04.004 49 2. Aaby P, Benn CS. Non-specific and sex-differential effects of routine vaccines: what evidence is needed to 50 take these effects into consideration in low-income countries? Human vaccines 2011;7(1):120-4. 51 52 [published Online First: 2011/02/01] 53 3. Stensballe LG, Ravn H, Birk NM, et al. BCG Vaccination at Birth and Rate of Hospitalization for Infection 54 Until 15 Months of Age in Danish Children: A Randomized Clinical Multicenter Trial. J Pediatric 55 Infect Dis Soc 2018;0(0):1-8. doi: 10.1093/jpids/piy029 [published Online First: 2018/04/11] 56 4. Martins CL, Benn CS, Andersen A, et al. A randomized trial of a standard dose of Edmonston-Zagreb 57 measles vaccine given at 4.5 months of age: effect on total hospital admissions. J Infect Dis 58 2014;209(11):1731-8. doi: 10.1093/infdis/jit804 [published Online First: 2014/01/18] 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 31 of 31 BMJ Open

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 For peer review only 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 45 on September 24, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from

Can earlier BCG vaccination reduce early infant mortality? Study protocol for a cluster randomised trial in Guinea- Bissau ForJournal: peerBMJ Open review only Manuscript ID bmjopen-2018-025724.R1

Article Type: Protocol

Date Submitted by the 14-Feb-2019 Author:

Primary Subject Global health Heading:

Secondary Subject Heading: Epidemiology, Paediatrics http://bmjopen.bmj.com/ BCG vaccine, Oral polio vaccine, Cluster randomised trial, Non-specific Keywords: effects of vaccines

on September 24, 2021 by guest. Protected copyright.

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 1 of 39 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Can earlier BCG vaccination reduce early infant mortality? Study protocol for a cluster 5 6 randomised trial in Guinea-Bissau 7 8 1,2,3 2,4 1 9 Sanne Marie Thysen , Aksel Karl Georg Jensen , Amabelia Rodrigues , Igualdino da Silva 10 Borges1, Peter Aaby1,2, Christine Stabell Benn1,2,5, Ane Bærent Fisker1,2,5 11 12 1 13 Bandim Health Project, Bissau, Guinea-Bissau 14 2Research Centre for Vitamins and Vaccines, Bandim Health Project, Statens Serum Institut, 15 16 Copenhagen, Denmark 17 3 18 Center for Global HealthFor (GloHAU), peer Aarhus review University, Aarhus, only Denmark 19 4 20 Section of Biostatistics, Department of Public Health, University of Copenhagen, Copenhagen, 21 Denmark 22 23 5OPEN, Institute of Clinical Research, University of Southern Denmark, Odense, Denmark 24 25 26 27 Corresponding author: Sanne Marie Thysen, Bandim Health Project, INDEPTH Network, 28 29 Apartado 861, 1004 Bissau Codex, Guinea-Bissau; [email protected] 30 31 32 33 34 35 36 37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 45 on September 24, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 2 of 39

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 3 of 39 BMJ Open

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 4 of 39

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 5 of 39 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Guinea-Bissau18. A major reason for this delay is the focus on reducing wastage of vaccines. 5 6 Freeze-dried vaccines like BCG have to be used within 6 hours after reconstitution with a diluent. 7 8 Therefore, a vial of BCG vaccine is often not opened unless 10-12 children are present for 9 10 vaccination. Hence, there are many missed vaccination opportunities. If BCG vaccine has profound 11 effects on neonatal mortality and morbidity many lives could be saved by providing BCG earlier. 12 13 14 In the present study, we aim to study the effect of BCG and OPV vaccinations provided at a single 15 home visit within 72 hours after birth on early infant mortality and morbidity among infants in a 16 17 rural and an urban setting in Guinea-Bissau. We hypothesise that BCG and OPV at birth provided at 18 For peer review only 19 a single home visit shortly after birth reduce early infant non-accidental mortality by 40%. 20 21 METHODS AND ANALYSES 22 23 Setting 24 25 The study is conducted in Bandim Health Project’s (BHP) Health and Demographic Surveillance 26 27 System (HDSS) sites in rural and urban Guinea-Bissau. 28 29 Rural site: The rural HDSS was established in 1990. The BHP teams survey women of fertile age 30 31 and their children below the age of 5 years in randomly selected clusters of villages in all the nine 32 health regions in the country. Children in the three rural regions closest to Bissau (Oio, Biombo and 33 34 Cacheu) are eligible for the study. 35 36 37 The rural clusters are followed with two-monthly visits by the mobile teams. At all visits, the http://bmjopen.bmj.com/ 38 women are asked whether they are pregnant. When a pregnancy is registered; the woman’s 39 40 nutritional status is assessed by measurement of a mid-upper-arm-circumference (MUAC); 41 42 information on antenatal care is collected prior to giving birth, as well as at the first visit after 43 delivery. Socio-economic factors (type of roofing, type of bathroom, possession of a mobile phone, 44 45 radio and generator) are registered. After the delivery, information on the place of delivery (home, on September 24, 2021 by guest. Protected copyright. 46 47 health facility) and who assisted the delivery is collected. 48 49 Urban site: The urban HDSS was established in 1978. Field assistants follow children below the 50 51 age of 3 years through home visits every third month. Pregnancies are registered every month and 52 53 followed monthly until birth. When a pregnancy is registered information on ethnic group, 54 gestational age, use of mosquito net is collected. After the delivery, information on place of delivery 55 56 (home, health facility) is collected. All children in the urban study area will be eligible for the study. 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 6 of 39

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Study nurses 5 6 Prior to study start, we selected study nurses to conduct home visits. In the rural study area, study 7 8 nurses were selected in collaboration with the regional health authorities. The nurses were selected 9 10 among nurses working at a health centre with a catchment area corresponding to the study clusters, 11 so the home visits can be done in addition to their normal routines. For the urban study area, the 12 13 BHP employed a study nurse fulltime. 14 15 Community Key Informants 16 17 In the rural areas, to ensure that the nurses are informed immediately about deliveries, we use 18 For peer review only 19 community key informants (CKI). CKIs were selected among residents in each rural village to 20 21 collect information about pregnancies, deliveries and deaths. The CKI communicate immediately 22 any delivery in the village to the study nurse. In the urban area, we take advantage of the close 23 24 follow-up by the field assistants, who circulate the zones daily. 25 26 Study design and randomisation 27 28 The trial is a cluster-randomised trial, randomising clusters (village clusters/sub-district zones) to 29 30 two different treatment groups, stratified by region and pre-trial mortality level (high/low). 31 32 Clusters, defined as village cluster in the rural area, and as sub-district zones in the urban area, were 33 34 randomised prior to study start using computer generated random numbers. All newborns are visited 35 36 as soon as possible after birth. In half of the clusters, we provide standard care and vaccines (BCG, 37 OPV) to children at the home visits; in the remaining clusters, children only receive standard care http://bmjopen.bmj.com/ 38 39 but no vaccines (Table 1). The trial is unblinded. 40 41 42 Inclusion criteria 43 All children registered during pregnancy are eligible for the study. Children can only be enrolled if 44 45 they are visited within 72 hours after birth. on September 24, 2021 by guest. Protected copyright. 46 47 Exclusion criteria 48 49 There are few exclusion criteria, because the study is expected to answer a pragmatic question about 50 51 the effect of BCG and OPV vaccination at home visits shortly after birth. 52 53  Children already BCG vaccinated 54 55  Moribund children (Expected not to survive the next 24 hours, as evaluated by the study nurse 56 57 at the enrolment visit) 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 7 of 39 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4  Children in rural villages where the BHP mobile teams coincidentally were in the village the 5 6 same day (and vaccinated all children) 7 8 9 Informed consent 10 11 All pregnant women receive an oral and a written explanation of the study, and are offered to 12 participate in the study at the time of registration of the pregnancy. In the rural area, the explanation 13 14 is given by the BHP nurse at the two-monthly visits. In the urban area, the explanation is given by 15 16 the fieldwork assistant at a monthly visit. It is explained that normally newborn children receive 17 their first vaccinations at the health centres. However, studies have suggested that it might be 18 For peer review only 19 beneficial to receive BCG and OPV early and the BHP will therefore test the effect of providing the 20 21 vaccines at home visits. All children receive a home visit with standard care. Children in half the 22 23 clusters (intervention clusters) receive BCG and OPV at enrolment. Children in the control clusters 24 can seek vaccination elsewhere and will be offered the BCG vaccine and OPV at the next home 25 26 visit by the BHP team if they did not get it elsewhere. If the pregnant woman accepts to take part in 27 28 the study, they are asked to sign or fingerprint a consent form. An independent witness confirms the 29 30 consent process by co-signing the form. 31 32 Enrolment 33 34 The study nurse visits every newborn child shortly after a CKI/mother calls, if possible on the same 35 day. At the home visit, the nurse asks the mother for confirmation of her consent for participation in 36 37 the study before revealing the randomisation. If the mother is not able to present the signed consent http://bmjopen.bmj.com/ 38 39 form, new consent forms are filled in. The nurse brings a questionnaire, a sticker with study number 40 41 and a vaccination card. 42 43 Intervention 44 45 For all children, the nurse examines and weighs the child, encourages skin-to-skin contact to keep on September 24, 2021 by guest. Protected copyright. 46 the newborn warm and if necessary performs umbilical cord care1. Following interview and 47 48 examination, group allocation is disclosed to the mother. For children in the control clusters, the 49 50 nurse informs about vaccination opportunities (vaccination at closest health centre or vaccination by 51 1 52 BHP nurse at next visit), as recommended by WHO . For children in the intervention clusters, the 53 nurse administers BCG and OPV (Table 1). 54 55 56 BCG vaccine 57 BCG vaccination is administered by intradermal injection; after vaccination most children develop a 58 59 scar at the injection site. Among BCG-vaccinated children, having a BCG scar is associated with 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 8 of 39

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 improved survival19-21. The proportion of children developing a scar after BCG vaccination depends 5 6 on the vaccination technique and strain22-24. All nurses have been trained prior to the trial and 7 8 supervised intensively during the beginning of the trial to ensure correct vaccination technique. The 9 10 Tokyo 172 BCG strain from BCG Japan is used for the study. The vaccine is prequalified by WHO 11 and is regularly used in the national vaccination programme in Guinea-Bissau. 12 13 14 OPV vaccine 15 The OPV vaccine is supplied from the national vaccination programme and thus strain, 16 17 manufacturer and batch may vary. For all vaccines, the manufacturer, batch and expiration date is 18 For peer review only 19 registered. 20 21 Follow-up 22 23 All children enrolled in the study receive two follow-up visits. At follow-up visits, the child is 24 25 weighed, and the BHP nurse examines lymphadenopathy and reaction at the vaccination site. 26 27 Follow-up in the rural regions 28 29 The study participants are followed through the rural HDSS, where all children below 5 years of 30 31 age are followed, and information on vital status, breastfeeding status, supplementary feeding, 32 MUAC, vaccinations, hospital admissions and whether the child has received interventions 33 34 provided in campaigns is collected. 35 36 The BHP teams are accompanied by a nurse, who administers vaccines. The BHP nurse performs 37 http://bmjopen.bmj.com/ 38 the follow-up visits for the present study. All children have their vaccination cards examined and 39 40 those missing one or more routine vaccines according to the schedule are offered these vaccines at 41 42 the follow-up visits. 43 44 Follow-up in the urban region 45 on September 24, 2021 by guest. Protected copyright. 46 The study participants receive two follow-up visits by the study nurse at 2 and 4 months of age. The 47 study nurse brings BCG and OPV for children in control clusters. For the remaining routine 48 49 vaccines as part of the national programme, the children are referred to the closest health centre. All 50 51 children below the age of 3 years are followed through home visits every third month, where 52 53 information on vital status, breastfeeding status, supplementary feeding, MUAC, vaccinations, 54 hospital admissions and vaccination campaigns are collected. 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 9 of 39 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Outcomes 5 6 The primary outcome is non-accidental mortality rate between enrolment visit or 24 hours after 7 8 birth, whichever comes last, and 60 days of life. Follow-up time will be censored at the first of visit 9 10 by the BHP, death due to accident, date of registration of first non-trial vaccine given after 11 enrolment, or migration. All children living in intervention and control clusters are followed 12 13 through the HDSS routines with collection of information on vital status for all children. Deaths are 14 15 classified as accidental or non-accidental based on information from verbal autopsies. 16 17 Secondary outcomes and potential effect modifiers 18 For peer review only 19 We will evaluate the effect on severe morbidity considered as the composite outcome of non- 20 21 accidental death and non-accidental hospital admission. With declining mortality, we may not be 22 able to obtain a conclusive result for the primary outcome. We have defined the sample size for the 23 24 composite outcome, but retained mortality as the primary outcome. 25 26 We will conduct secondary analyses to assess whether an effect vary by low birthweight (<2500g), 27 sex, maternal BCG scarring, and season to gain a better knowledge of the potential effect modifiers 28 29 (a list of all outcomes can be found in Table 2). 30 31 We will study the cost-effectiveness of providing BCG and OPV at a home visit using the effects on 32 33 mortality and hospital admission from the present project. 34 35 Data management 36 37 All data is collected on paper forms and entered in the BHP office. Records with registration of all http://bmjopen.bmj.com/ 38 pregnant women are copied from the data tables of the routine HDSS to the BCG trial data table. 39 40 Forms for follow-up are pre-printed with identification information of the enrolled children. For 41 42 routine HDSS visits, a list of all pregnant women is printed prior to the visit to ensure that all pregnant 43 44 women are invited to participate in the study prior to giving birth. Following data entry, the data is 45 checked for consistency. Inconsistencies between the BCG trial data and the HDSS data are checked. on September 24, 2021 by guest. Protected copyright. 46 47 Main outcome events are reviewed individually. 48 49 50 Statistical analyses 51 General analysis principles applied in all analyses can be found in appendix 1. In Cox-proportional 52 53 hazards models with age as underlying time-scale, we will assess the effect of providing BCG and 54 55 OPV at a single home visit. We will allow for different baseline hazards according to the factors 56 used in the randomisation (region and pre-study mortality level) and sex. Children enter the analysis 57 58 on the date of enrolment or 24 hours of age, whichever comes last, and remain in the analysis until 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 10 of 39

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 the first subsequent visit by the BHP team, date of registering first non-trial vaccine after enrolment, 5 6 death or migration within the first 60 days of life. In secondary analyses, we will investigate 7 8 whether the effect of BCG+OPV differ by the following potential effect modifiers, which in prior 9 10 trials have been important determinants of the effect: low birthweight (<2500 g), sex, maternal 11 BCG scarring, and season. Further details are provided in appendix 2. 12 13 14 Secondary outcomes are non-accidental hospital admissions, severe morbidity (composite outcome 15 of non-accidental mortality and non-accidental hospital admissions), consultations, growth (mid- 16 17 upper-arm circumference and weight-for-age z-score), and BCG scarring. Analyses are described in 18 For peer review only 19 appendix 2. We will furthermore perform a cost effectiveness analysis seeking to measure the cost 20 21 per death averted using a societal perspective, contrasting the costs of vaccine provision in the 22 present programme and an outreach system as tested in the trial. The final analysis plan is deposited 23 24 with the Data Safety and Monitoring Board. A list of all planned analyses is provided in appendix 2. 25 26 Sample size considerations 27 28 We estimated the design effect to 1.35 (ratio of square of the standard errors for the cluster- 29 30 adjusted/unadjusted HRs) based on pre-trial event data (deaths and hospitalisation) from the rural 31 32 HDSS clusters, where the study is conducted. We base our sample size calculation on the combined 33 secondary outcome for feasibility reasons, but retain mortality as the primary outcome. We 34 35 estimated the proportion of events to be 2.4% in the absence of our intervention (prior data). Thus, 36 37 in order to obtain 80% power to detect a reduction in early infant severe morbidity if the true http://bmjopen.bmj.com/ 38 reduction of BCG and OPV provided at home visits is larger than 40%, we will need to enrol at 39 40 least 6666 children. 41 42 43 Time schedule 44 45 The trial was initiated with a pilot study in Biombo region in June 2015. The trial was scaled up to on September 24, 2021 by guest. Protected copyright. 46 full study including Oio and Cacheu regions in July 2016. The trial was expanded to include the 47 48 urban study area in July 2017. We expect enrolments to end in December 2020. 49 50 51 ETHICS AND DISSEMINATION 52 53 Ethical considerations 54 55 BCG is recommended at birth but vaccination is often delayed; in rural Guinea-Bissau most 56 18 57 children receive BCG after 1 month . While there is now increasing acceptance that BCG may 58 have beneficial non-specific effects, we are still a long way from altering the implementation of the 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 11 of 39 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 vaccination programme. The proposal compares two ways of delivering BCG and OPV; both are an 5 6 improvement relative to the current situation in rural Guinea-Bissau, where less than 40% of all 7 8 infants get BCG during the first month of life25. Hence, no child receives BCG later than it would 9 10 have done, had the trial not been carried out. The trial is approved by the Guinean Ethical 11 Committee (Reference number: 0016/CNES/INASA/2015) and the Danish Ethics Committee has 12 13 given its consultative approval. The trial is registered at clinicaltrials.gov (Clinicaltrials.gov ID: 14 15 NCT02504203). 16 17 Safety monitoring 18 For peer review only 19 A Data Safety and Monitoring Board has been formed consisting of a paediatrician (Anja Poulsen, 20 21 Rigshospitalet, Denmark), a statistician (Morten Frydenberg, Aarhus University, Denmark) and an 22 epidemiologist (Torben Sigsgaard, Aarhus University, Denmark). The members have been 23 24 appointed on their experience, reputation for objectivity, absence of conflicts of interest, and 25 26 knowledge on clinical trial methodology. 27 28 The vaccines used for the trial are prequalified by WHO to be given at birth. Adverse reactions are 29 30 rare for both BCG and OPV. At the two first visits after enrolment in the study, the BHP nurse 31 32 examines the BCG vaccination site and the axillary lymph glands of all children to assess 33 suppurative lymphadenitis as an adverse reaction to the BCG vaccination. Other serious adverse 34 35 events are captured through primary and/or secondary outcomes (mortality, hospitalisations and 36 37 consultations). http://bmjopen.bmj.com/ 38 39 Dissemination 40 41 42 The results of the study will be published in international peer-reviewed journals and results will be 43 communicated to the national institute of public health in Guinea-Bissau. After publication of the 44 45 main results on completion of the trial, data requests can be submitted to the researchers at Bandim on September 24, 2021 by guest. Protected copyright. 46 47 Health Project. 48 49 Protocol amendments 50 51 52 Any protocol modifications will be discussed with the Data Safety and Monitoring Board, and 53 changes will be added to the trial registration. 54 55 56 Patient and public involvement 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 12 of 39

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 The communities were involved in locating households, when the HDSS was setup and contributed 5 6 information allowing tracing of internal migrants between villages throughout the study period. No 7 8 participants were involved in setting the research question or the outcome measure, nor were they 9 10 involved in developing plans for recruitment, design, or implementation of the study. Local 11 community key informants were selected to ensure timely information about birth. Nurses from 12 13 local health centres are involved in the study and have been trained to perform enrolment. This 14 15 ensures a close collaboration with the local health system. No participant was asked to advise on 16 interpretation or writing up the results. The results are disseminated to the national public health 17 18 institute. There are no Forplans to disseminatepeer thereview results of the researchonly to study participants. Study 19 20 results will be disseminated to local nurses involved in the study. 21 22 23 24 25 DISCUSSION 26 27 WHO recommends BCG and OPV at birth. However, due to the focus on not wasting vaccine 28 29 doses, BCG vaccination is often delayed18 26-28. Previous trials from Guinea-Bissau have showed 30 6 7 9 31 that early vaccination may have major impact on neonatal mortality among low-weight children , 32 and observational studies have suggested that BCG also has beneficial NSEs among normal-birth- 33 34 weight children3 11. In 2014, the Strategic Advisory Group of Experts on Immunization reviewed 35 36 the non-specific effects of the BCG vaccine, and concluded that the evidence was consistent with 37 beneficial NSEs. However, the bulk of evidence came from observational studies and they therefore http://bmjopen.bmj.com/ 38 39 called for randomised trials. To our knowledge, this will be the first randomised trial assessing the 40 41 NSEs of BCG and OPV vaccination among children in a rural low-income setting. The SAGE 42 43 review did not include OPV, but a recent trial from Guinea-Bissau randomised children to BCG 44 only or BCG and OPV, and found that OPV was associated with a 32% reduction in mortality29. 45 on September 24, 2021 by guest. Protected copyright. 46 Thus, both BCG and OPV may have beneficial NSEs. The results will therefore be very important 47 48 in broadening our understanding of the NSEs of BCG and OPV at birth. We will however not be 49 50 able to distinguish the effects of BCG and OPV in the present trial. 51 52 WHO recommends three home visits after birth, which is resource demanding and not implemented 53 54 in many low-income countries. It is not recommended that the home visits are utilised for early 55 vaccination. However, we hypothesise that a single home visit after birth will have major effects on 56 57 early infant mortality if BCG and OPV vaccination are added to the home visit. If BCG and OPV 58 59 provided at a single home visit can reduce early infant mortality, this should be included in the 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 13 of 39 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 WHO recommendations. A single home visit with vaccines will be easier to implement in a strained 5 6 health system than three home visits within 7 days. In countries, where home visits are already in 7 8 place, vaccines can easily be added to reduce early infant mortality. 9 10 11 12 13 14 15 16 17 18 For peer review only 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 45 on September 24, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 14 of 39

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 15 of 39 BMJ Open

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 16 of 39

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Authors’ contributions 5 6 AR, PA and CB conceived the idea for the trial. SMT, PA, CB and ABF designed the study. SMT 7 8 set up the study with help from ABF. SMT, IDSB and ABF supervised data collection. SMT, 9 10 AKGJ, and ABF planned the data analyses. SMT drafted the manuscript and data analysis plan with 11 help from AKGJ and ABF. All authors read and approved the final manuscript and data analysis 12 13 plan. 14 15 16 Acknowledgements 17 We thank our Data Safety and Monitoring board, Anja Poulsen, Morten Frydenberg, and Torben 18 For peer review only 19 Sigsgaard, for insightful comments on planned analysis, study design and implementation. 20 21 22 Funding 23 24 Karen Elise Jensen foundation is the main funder of the trial. Laerdals Foundation supports the trial. 25 The Research Centre for Vitamins and Vaccines, Bandim Health Project receives support from the 26 27 Danish National Research Foundation [DNRF108]. The funders had no role in the design of the 28 29 study or writing of the protocol. 30 31 32 Competing interests 33 The authors declare no competing interests. 34 35 Access to data 36 37 Data is available on a collaborative basis. http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 45 on September 24, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from

Page 17 of 39 BMJ Open

1 2 3 4 Figure 1: Recommended vaccination schedule in Guinea-Bissau 5 6 7 8 9 10 11 12 For peer review only 13 14 15 16

17 http://bmjopen.bmj.com/ 18 19 20 21 22 23 24

25 on September 24, 2021 by guest. Protected copyright. 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from

BMJ Open Page 18 of 39

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Table 2: Outcomes 5 6 Primary outcome 7 1 8  Early infant non-accidental mortality 9 Potential effect modifiers 10  Low birthweight defined as birthweight lower than 11 2500g 12  Sex 13 14  Maternal BCG scarring 15  Season 16 Secondary outcomes 17  Non-accidental hospital admission1 18 For peer review only 19  Severe morbidity (composite outcome of non-accidental 1 20 mortality and non-accidental hospital admissions) 21  Consultations2 22  Growth3 23  Mid-upper-arm circumference 24 25  Weight-for-age z-score 26  BCG scarring2 27  Cost-effectiveness of providing BCG and OPV at home 28 visits 29 1 Analysed using Cox proportional hazards models. Further details are provided in Appendix 2. 30 2 31 Analysed using Log-binomial regression. Further details are provided in Appendix 2. 32 3 Analysed using Linear regression. Further details are provided in Appendix 2. 33 34 35 36 37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 45 on September 24, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 20 of 39 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For peer review only 17 18 19 20 21 22 Figure 1: Recommended vaccination schedule in Guinea-Bissau 23 24 25 26 27 28 29 30 31 32

33 http://bmjopen.bmj.com/ 34 35 36 37 38 39 40

1 2

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 22 of 39

1 2

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 23 of 39 BMJ Open

1 2

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 24 of 39

1 2

18 For peer- 60 days review only 19 - Death due to accident 20 - Date of registering first non-trial vaccine given after 21 22 enrolment 23 - Migration (migration out of the study area as per HDSS 24 definition) 25 26 Time scale Age 27 Failure definition Death 28 29 Statistical tool Cox proportional hazards model 30 Stratification We will employ models that allow for different baseline hazards 31 according to sex, region and pre-study mortality level 32 33 Clustering We will use robust standard errors to account for clustering within 34 village cluster 35 Outline stata code 36 37 For analysis: stset outdate, f(dead==1) enter (max(dob+1, date_enrol) /// http://bmjopen.bmj.com/ 38 exit (min ( dnasc+60, date_mobileteam) origin(dob) 39 40 41 stcox random, strata(sex reg prmorlev) vce(cl regam) 42 43 For model checking: estat phtest, detail† 44 45 stphplot, strata(random) adj(sex reg prmorlev) on September 24, 2021 by guest. Protected copyright. 46 stcox random, strata(sex reg prmorlev) vce(cl regam) tvc(random) 47 texp(_t) 48 49 50 *reg=region, prmorlev=pre-study mortality level, regam=village 51 cluster 52 † If the proportional hazards assumption is not fulfilled we will still report the marginal hazard ratio 53 54 but it will be supplemented by hazard ratios for two-three categorized time periods. 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 25 of 39 BMJ Open

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 3 Effect-modifier analyses of primary outcome 5 6 We will assess whether the effect of the intervention on the primary effect measure is modified by 7 8 the following potential effect modifiers. 9 10 Table 3. Sex as a potential effect modifier of the primary outcome 11 12 Potential effect Sex 13 modifier 14 15 Design We will perform the analysis as describe above (for the primary 16 analysis) allowing the effect of the intervention to differ between the 17 sexes. 18 For peer review only 1 2 19 Reasoning Previous studies have found sex-differential non-specific effects , 20 therefore, we will assess the sex-differential effects. 21 Outline stata code: 22 23 For analysis: stcox random#sex sex, strata(sex reg prmorlev) vce(cl regam) 24 contrast random#sex 25 26 27 For model checking: estat phtest, detail† 28 stphplot, strata(random_sex) adj(sex reg prmorlev) 29 stcox random#sex sex, strata(sex reg prmorlev) vce(cl regam) /// 30 31 tvc(random#sex sex) texp(_t) 32 33 *reg=region, prmorlev=pre-study mortality level, regam=village 34 35 cluster, random_sex=a four-level variable based on the four possible 36 combinations of random and sex 37 † If the proportional hazards assumption is not fulfilled we will still report the marginal hazard ratio http://bmjopen.bmj.com/ 38 but it will be supplemented by hazard ratios for two-three categorized time periods. 39 40 41 Table 4. Maternal BCG scar as a potential effect modifier of the primary outcome 42 43 Potential effect Maternal BCG scar (yes/no) 44 modifier 45 on September 24, 2021 by guest. Protected copyright. 46 Design We will perform the analysis as described above (for the primary 47 analysis) allowing the effect of the intervention to differ by maternal 48 BCG-scar status 49 50 Reasoning A recent randomized trial in Denmark found that the effect of BCG 51 varied by whether the mother had received BCG or not3. Since, BCG 52 scar is a life-long marker of a successful BCG-vaccination, we will 53 54 assess whether the effect of BCG differs by maternal BCG-scar status 55 Outline stata code: 56 For analysis: stcox random#mBCGscar mBCGscar, strata(sex reg prmorlev) vce(cl 57 58 regam) 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 26 of 39

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 contrast random#mBCGscar 5 6 For model checking: 7 estat phtest, detail† 8 stphplot, strata(random_mBCGscar) adj(sex reg prmorlev) 9 10 stcox random#mBCGscar mBCGscar, strata(sex reg prmorlev) /// 11 vce(cl regam) tvc(random#mBCGscar mBCGscar) texp(_t) 12 13 14 * mBCGscar= maternal BCG scar, reg=region, prmorlev=pre-study 15 mortality level, regam=village cluster, random_mBCGscar=a four- 16 level variable based on the four possible combinations of random and 17 18 Formaternal peer BCG scar review only 19 † If the proportional hazards assumption is not fulfilled we will still report the marginal hazard ratio 20 but it will be supplemented by hazard ratios for two-three categorized time periods. 21 22 Table 5. Low birthweight as a potential effect modifier of the primary outcome 23 24 Potential effect Low birthweight (<2500g: yes/no) 25 modifier 26 27 Design We will perform the analysis as described above (for the primary 28 analysis) allowing the effect of the intervention to differ by birthweight 29 strata 30 31 Reasoning Birthweight is an important risk factor for mortality. Previous 32 randomised trials from Guinea-Bissau assessing the effect of BCG on 33 mortality have been performed among low-birth-weight children. 34 35 Outline stata code: 36 For analysis: stcox random#LBW LBW, strata(sex reg prmorlev) vce(cl regam) 37 contrast random#LBW http://bmjopen.bmj.com/ 38 39 40 For model checking: estat phtest, detail† 41 stphplot, strata(random_LBW) adj(sex reg prmorlev) 42 43 stcox random#LBW LBW, strata(sex reg prmorlev) vce(cl regam) /// 44 tvc(random#LBW LBW) texp(_t) 45 on September 24, 2021 by guest. Protected copyright. 46 47 * LBW=Low birthweight, reg=region, prmorlev=pre-study mortality 48 level, regam=village cluster, random_LBW=a four-level variable 49 based on the four possible combinations of random and low 50 51 birthweight 52 † If the proportional hazards assumption is not fulfilled we will still report the marginal hazard ratio 53 but it will be supplemented by hazard ratios for two-three categorized time periods. 54 55 56 Table 6. Season as a potential effect modifier of the primary outcome 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 27 of 39 BMJ Open

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Potential effect Season of birth (Dry: December-May/Rainy: June-November) 5 6 modifier 7 Design We will perform the analysis as described above (for the primary 8 analysis) allowing the effect of the intervention to differ by season of 9 10 birth 11 Reasoning Previous studies have found that the effect of some vaccines is stronger 12 in the dry season4. Therefore, we would like to assess if the effect of 13 14 BCG and OPV differs according to season. 15 Outline stata code: 16 For analysis: stcox random#season season, strata(sex reg prmorlev) vce(cl regam) 17 18 Forcontrast peer random#season review only 19 20 For model checking: estat phtest, detail† 21 22 stphplot, strata(random_season) adj(sex reg prmorlev) 23 stcox random#season season, strata(sex reg prmorlev) vce(cl regam) /// 24 tvc(random#season season) texp(_t) 25 26 27 *reg=region, prmorlev=pre-study mortality level, regam=village 28 cluster, random_season=a four-level variable based on the four 29 30 possible combinations of random and season 31 † If the proportional hazards assumption is not fulfilled we will still report the marginal hazard ratio 32 but it will be supplemented by hazard ratios for two-three categorized time periods. 33 34 35 4 Primary analyses of secondary outcomes 36 37

Non-accident morbidity http://bmjopen.bmj.com/ 38 39 Since it can be difficult to distinguish between hospital admissions and outpatient contact through 40 41 interviews, we have defined hospital admissions as an overnight stay in a health facility. Hospital 42 admissions due to accidents are ignored but the follow-up time is (interval) censored while the child 43 44 is admitted. 45 on September 24, 2021 by guest. Protected copyright. 46 Table 7: Non-accident hospitalisation 47 Population Identical to primary analysis of primary outcome 48 49 Observation period From: Enrolment visit or 24 hours after birth, whichever comes last 50 To: 60 days of life 51 52 Censoring, first of: 53 - Visit by the BHP 54 - Date of registering first non-trial vaccine given after 55 56 enrolment 57 - Death 58 - Hospital admission due to accident 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 28 of 39

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 - 60 days 5 6 - Migration (migration out of the study area as per HDSS 7 definition) 8 Time scale Age 9 10 Failure definition First hospital admission – only overnight hospitalisations or arrival at 11 the hospital and death within the first day will be considered in this 12 13 analysis. 14 Statistical tool Cox proportional hazards model 15 16 Stratification We will employ models that allow for different baseline hazards 17 according to sex, region and pre-study mortality level 18 Clustering ForWe peer will use robust review standard errors only to account for clustering within 19 20 village cluster 21 Outline stata code 22 For analysis: stset outdate, f(hosp==1) enter (max(dob+1, date_enrol) /// 23 24 exit (min ( dnasc+60, date_mobileteam) origin(dob) 25 26 stcox random, strata(sex reg prmorlev) vce(cl regam) 27 28

29 For model checking: estat phtest, detail† 30 stphplot, strata(random) adj(sex reg prmorlev) 31 stcox random, strata(sex reg prmorlev) vce(cl regam) tvc(random) 32 33 texp(_t) 34 35 *reg=region, prmorlev=pre-study mortality level, regam=village 36 37 cluster http://bmjopen.bmj.com/ 38 † If the proportional hazards assumption is not fulfilled we will still report the marginal hazard ratio 39 but it will be supplemented by hazard ratios for two-three categorized time periods. 40 41 Severe morbidity 42 43 We will evaluate the effect on severe morbidity considered as the composite outcome of non- 44 45 accidental death and non-accidental hospital admission. Since it can be difficult to distinguish on September 24, 2021 by guest. Protected copyright. 46 47 between hospital admissions and outpatient contact through interviews, we have defined hospital 48 admissions as an overnight stay in a health facility. Hospital admissions due to accidents are 49 50 ignored but the follow-up time is (interval) censored while the child is admitted. The potential effect 51 52 modifiers for the primary outcome specified in tables 3-6 will also be assessed for the composite 53 54 outcome. 55 Table 8: Severe morbidity 56 57 Population Identical to primary analysis of primary outcome 58 Observation period From: Enrolment visit or 24 hours after birth, whichever comes last 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 29 of 39 BMJ Open

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 To: 60 days of life 5 6 Censoring, first of: 7 - Visit by the BHP 8 - Date of registering first non-trial vaccine given after 9 10 enrolment 11 - Death/Hospital admission due to accident 12 - 60 days 13 14 - Migration (migration out of the study area as per HDSS 15 definition) 16 Time scale Age 17 18 Failure definition For Death peer or first hospitalreview admission only 19 20 Statistical tool Cox proportional hazards model 21 Stratification We will employ models that allow for different baseline hazards 22 according to sex, region and pre-study mortality level 23 24 Clustering We will use robust standard errors to account for clustering within 25 village cluster 26 Outline stata code: 27 28 For analysis stset outdate, f(event==1) enter (max(dob+1, date_enrol) /// 29 exit (min ( dnasc+60, date_mobileteam) origin(dob) 30 31 32 stcox random, strata(sex reg prmorlev) vce(cl regam) 33 34 For model checking estat phtest, detail† 35 stphplot, strata(random) adj(sex reg prmorlev) 36 37 stcox random, strata(sex reg prmorlev) vce(cl regam) tvc(random) http://bmjopen.bmj.com/ 38 texp(_t) 39 40 41 *reg=region, prmorlev=pre-study mortality level, regam=village 42 cluster 43 † If the proportional hazards assumption is not fulfilled we will still report the marginal hazard ratio 44 45 but it will be supplemented by hazard ratios for two-three categorized time periods. on September 24, 2021 by guest. Protected copyright. 46 47 48 Table 9: All-cause consultations 49 Population Identical to primary analysis of primary outcome 50 51 Observation period From: Enrolment visit or 24 hours after birth, whichever comes last 52 To: 60 days of life 53 Censoring, first of: 54 55 - Visit by the BHP 56 - Date of registering first non-trial vaccine given after 57 enrolment 58 59 - 60 days 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 30 of 39

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 - Migration (migration out of the study area as per HDSS 5 6 definition) 7 - Death 8 - End of study 9 10 Failure definition An out-patient consultation within the observation period 11 Statistical tool Log-binomial regression 12 13 Adjustment We will adjust the analysis for sex, region and pre-study mortality 14 level 15 Clustering We will use robust standard errors to account for clustering within 16 17 village cluster 18 Stata code ForBinreg peer cons random review sex b1.reg prmorlev,only rr vce(cl regam) 19 *cons=out-patient consultation, reg=Region, prmorlev=Pre-study 20 21 mortality level, regam=village cluster 22 23 24 Growth 25 26 Table 10: Mid-upper-arm circumference (MUAC) 27 Population Identical to primary analysis of primary outcome 28 29 Observation time point First visit by the mobile teams 30 Growth measures MUAC will be analysed using the measured value 31 Statistical tool Linear regression 32 33 Adjustment We will adjust the analysis for MUAC at enrolment, age at MUAC 34 measurement, sex, region and pre-study mortality level 35 Clustering We will use robust standard errors to account for clustering within 36 37 village cluster http://bmjopen.bmj.com/ 38 Stata code Regress MUAC random sex b1.reg prmorlev MUACenrol 39 MUACage, /// vce(cl regam) 40 41 *reg=region, prmorlev=pre-study mortality level, 42 MUACenrol=MUAC at enrolment, MUACage= age at MUAC 43 assessment, regam=village cluster 44 45 on September 24, 2021 by guest. Protected copyright. 46 47 Table 11: Weight-for-age z-score 48 Population Identical to primary analysis of primary outcome 49 50 Observation time point First visit by the mobile teams 51 Growth measures Weight will be analysed using the WHO weight-for-age z-score 52 Statistical tool Linear regression 53 54 Adjustment We will adjust the analysis for weight-for-age at enrolment, sex, 55 region and pre-study mortality level 56 Clustering We will use robust standard errors to account for clustering within 57 58 village cluster 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 31 of 39 BMJ Open

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Stata code Regress w-z-score random w-z-enrol sex b1.reg prmorlev, vce(cl 5 6 regam) 7 *w-z-score=weight-for-age z-score at first visit by the mobile teams, 8 w-z-enrol=weight-for-age z-score at enrolment, reg=region, 9 10 prmorlev=pre-study mortality level, regam=village cluster 11 12 13 BCG scarring 14 15 Table 12: BCG scarring 16 17 Population Identical to primary analysis of primary outcome 18 Observation timepointFor First peer visit after review6 months of age only 19 Failure definition Scar (yes/no) 20 21 Statistical tool Log-binominal regression 22 Adjustment We will adjust the analysis for sex, region and pre-study mortality 23 level 24 25 Clustering We will use robust standard errors to account for clustering within 26 village cluster 27 Stata code Binreg scar random sex b1.reg prmorlec, rr vce(cl regam) 28 29 *reg=region, prmorlev=pre-study mortality level, regam=village 30 cluster 31 32 33 Cost-effectiveness of providing BCG and OPV at birth 34 35 A cost effectiveness analysis seeking to measure the cost per death averted using a societal 36 37 perspective will be performed, contrasting the costs of vaccine provision in the present programme http://bmjopen.bmj.com/ 38 39 and an outreach system as tested in the trial. The costs/savings associated with different rates of 40 consultations and admissions will also be taken into account. 41 42 43 Suppurative lymphadenitis 44 45 We will assess the incidence of suppurative lymphadenitis as a reaction to BCG vaccination in the on September 24, 2021 by guest. Protected copyright. 46 intervention and control clusters. Other serious adverse events to the BCG and OPV vaccine will be 47 48 captured through the outcome measures (mortality, hospital admission and consultations). 49 50 5 Sensitivity analyses to test for robustness of conclusions 51 52 Table 13: Cause-specific death 53 54 Population Identical to primary analysis of primary outcome 55 Observation period From: Enrolment visit or 24 hours after birth, whichever comes last 56 57 To: 60 days of life 58 Censoring, first of: 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 32 of 39

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 - Visit by the BHP 5 6 - 60 days 7 - Death due to accident 8 - Date of registering first non-trial vaccine given after 9 10 enrolment 11 - Migration (migration out of the study area as per HDSS 12 definition) 13 14 Time scale Age 15 Failure definition Death due to: 16 Malaria, Respiratory Infection, Sepsis, Gastrointestinal disease, Other 17 18 Statistical tool ForCox peer proportional review hazards model only 19 Stratification We will employ models that allow for different baseline hazards 20 21 according to sex, region and pre-study mortality level 22 Clustering We will use robust standard errors to account for clustering within 23 village cluster 24 25 Outline stata code 26 For analysis: stset outdate, f(event==1&cause==X) enter (max(dob+1, date_enrol) 27 /// exit (min (dnasc+60, date_mobileteam) origin(dob) 28 29 30 stcox random, strata(sex reg prmorlev) vce(cl regam) 31 32 33 For model checking: estat phtest, detail† 34 stphplot, strata(random) adj(sex reg prmorlev) 35 stcox random, strata(sex reg prmorlev) vce(cl regam) tvc(random) 36 texp(_t) 37 http://bmjopen.bmj.com/ 38 39 *reg=region, prmorlev=pre-study mortality level, regam=village 40 cluster 41 42 † If the proportional hazards assumption is not fulfilled we will still report the marginal hazard ratio 43 but it will be supplemented by hazard ratios for two-three categorized time periods. 44 45 on September 24, 2021 by guest. Protected copyright. 46 In sensitivity analyses, we will furthermore, assess whether the conclusions are robust to the 47 48 following: 49 50 - Censoring follow-up at general health intervention campaigns (e.g. OPV campaigns) 51 52 - Altering the population to using an intention-to-treat approach, including all children, who 53 54 had a home visit by a study nurse (i.e., including children who did not receive assigned 55 treatment, children who were not enrolled because they were moribund, or who did not 56 57 accept to participate). 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 33 of 39 BMJ Open

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 - Altering the outcome from non-accidental mortality to all-cause mortality including also 5 6 deaths due to accidents. 7 8 - For the non-accident hospitalisations, we will perform the analyses allowing for repeated 9 10 hospitalisations. A child will return to the at-risk population the day after discharge from the 11 hospital. 12 13 14 References 15 16 1. Benn CS, Netea MG, Selin LK, et al. A small jab - a big effect: nonspecific immunomodulation by vaccines. 17 Trends in immunology 2013;34(9):431-9. doi: 10.1016/j.it.2013.04.004 18 For peer review only 19 2. Aaby P, Benn CS. Non-specific and sex-differential effects of routine vaccines: what evidence is needed to 20 take these effects into consideration in low-income countries? Human vaccines 2011;7(1):120-4. 21 [published Online First: 2011/02/01] 22 3. Stensballe LG, Ravn H, Birk NM, et al. BCG Vaccination at Birth and Rate of Hospitalization for Infection 23 Until 15 Months of Age in Danish Children: A Randomized Clinical Multicenter Trial. J Pediatric 24 Infect Dis Soc 2018;0(0):1-8. doi: 10.1093/jpids/piy029 [published Online First: 2018/04/11] 25 4. Martins CL, Benn CS, Andersen A, et al. A randomized trial of a standard dose of Edmonston-Zagreb 26 measles vaccine given at 4.5 months of age: effect on total hospital admissions. J Infect Dis 27 28 2014;209(11):1731-8. doi: 10.1093/infdis/jit804 [published Online First: 2014/01/18] 29 30 31 32 33 34 35 36 37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 45 on September 24, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 34 of 39

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 5 6 7 8 SPIRIT 2013 Checklist: Recommended items to address in a clinical trial protocol and 9 related documents* 10 11 Section/item Item Description 12 No 13 14 Administrative information 15 16 Title 1 Descriptive title identifying the study design, population, interventions, 17 and, if applicable, trial acronym 18 ForPage peer 1, lines 1-2review only 19 20 Trial registration 2a Trial identifier and registry name. If not yet registered, name of 21 22 intended registry 23 Page 2, lines 25-26. 24 25 2b All items from the World Health Organization Trial Registration Data 26 Set 27 28 Protocol version 3 Date and version identifier 29 30 Funding 4 Sources and types of financial, material, and other support 31 Page 15, lines 7-11. 32 33 Roles and 5a Names, affiliations, and roles of protocol contributors 34 responsibilities Page 1, lines 3-11 and page 15, lines 1-6 35 36 5b Name and contact information for the trial sponsor 37 http://bmjopen.bmj.com/ 38 There is no trial sponsor 39 40 5c Role of study sponsor and funders, if any, in study design; collection, 41 management, analysis, and interpretation of data; writing of the report; 42 and the decision to submit the report for publication, including whether 43 they will have ultimate authority over any of these activities 44 45 Page 15, lines 10-11. on September 24, 2021 by guest. Protected copyright. 46 47 5d Composition, roles, and responsibilities of the coordinating centre, 48 steering committee, endpoint adjudication committee, data 49 management team, and other individuals or groups overseeing the 50 trial, if applicable (see Item 21a for data monitoring committee) 51 52 Page 15, lines 1-6 and page 11, lines 3-14 53 54 Introduction 55 56 Background and 6a Description of research question and justification for undertaking the 57 rationale trial, including summary of relevant studies (published and 58 unpublished) examining benefits and harms for each intervention 59 60 Pages 1-2

1 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 35 of 39 BMJ Open

1 2 6b Explanation for choice of comparators 3 Pages 1-2 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 5 Objectives 7 Specific objectives or hypotheses 6 Page 2, lines 8-9. 7 8 Trial design 8 Description of trial design including type of trial (eg, parallel group, 9 crossover, factorial, single group), allocation ratio, and framework (eg, 10 11 superiority, equivalence, noninferiority, exploratory) 12 Pages 6-10 13 14 Methods: Participants, interventions, and outcomes 15 16 Study setting 9 Description of study settings (eg, community clinic, academic hospital) 17 18 Forand peer list of countries review where data onlywill be collected. Reference to where 19 list of study sites can be obtained 20 Pages 5-6 21 22 Eligibility criteria 10 Inclusion and exclusion criteria for participants. If applicable, eligibility 23 criteria for study centres and individuals who will perform the 24 25 interventions (eg, surgeons, psychotherapists) 26 Page 6, line 21 to page 7, line 2 27 28 Interventions 11a Interventions for each group with sufficient detail to allow replication, 29 including how and when they will be administered 30 Page 7, lines 22-27 31 32 11b Criteria for discontinuing or modifying allocated interventions for a 33 given trial participant (eg, drug dose change in response to harms, 34 35 participant request, or improving/worsening disease) 36 Not applicable as the intervention is only provided once at enrolment. 37 http://bmjopen.bmj.com/ 38 11c Strategies to improve adherence to intervention protocols, and any 39 procedures for monitoring adherence (eg, drug tablet return, 40 laboratory tests) 41 42 Not applicable as the intervention is only provided once at enrolment 43 44 11d Relevant concomitant care and interventions that are permitted or 45 prohibited during the trial on September 24, 2021 by guest. Protected copyright. 46 Not applicable as the intervention is only provided once at enrolment 47 48 Outcomes 12 Primary, secondary, and other outcomes, including the specific 49 measurement variable (eg, systolic blood pressure), analysis metric 50 51 (eg, change from baseline, final value, time to event), method of 52 aggregation (eg, median, proportion), and time point for each 53 outcome. Explanation of the clinical relevance of chosen efficacy and 54 harm outcomes is strongly recommended 55 56 Page 8 line 28 to page 9 line 14 and page 17 table 2 and Appendix 2. 57 58 59 60

2 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 36 of 39

1 2 Participant 13 Time schedule of enrolment, interventions (including any run-ins and 3 timeline washouts), assessments, and visits for participants. A schematic BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 diagram is highly recommended (see Figure) 5 6 Page 16, Table 1 7 8 Sample size 14 Estimated number of participants needed to achieve study objectives 9 and how it was determined, including clinical and statistical 10 assumptions supporting any sample size calculations 11 Page 10, lines 10-16 12 13 Recruitment 15 Strategies for achieving adequate participant enrolment to reach 14 target sample size 15 16 Page 6, lines 1-12 17 18 Methods: AssignmentFor of interventionspeer review (for controlled trials)only 19 20 Allocation: 21 22 Sequence 16a Method of generating the allocation sequence (eg, computer- 23 generation generated random numbers), and list of any factors for stratification. 24 To reduce predictability of a random sequence, details of any planned 25 restriction (eg, blocking) should be provided in a separate document 26 27 that is unavailable to those who enrol participants or assign 28 interventions 29 Page 6, lines 16-20 30 31 Allocation 16b Mechanism of implementing the allocation sequence (eg, central 32 concealment telephone; sequentially numbered, opaque, sealed envelopes), 33 34 mechanism describing any steps to conceal the sequence until interventions are 35 assigned 36 Page 7, lines 16-27 37 http://bmjopen.bmj.com/ 38 Implementation 16c Who will generate the allocation sequence, who will enrol participants, 39 and who will assign participants to interventions 40 41 Page 6, lines 16-20 42 Blinding 17a Who will be blinded after assignment to interventions (eg, trial 43 44 (masking) participants, care providers, outcome assessors, data analysts), and 45 how on September 24, 2021 by guest. Protected copyright. 46 Page 6, line 20. 47 48 17b If blinded, circumstances under which unblinding is permissible, and 49 procedure for revealing a participant’s allocated intervention during 50 51 the trial 52 Not applicable 53 54 Methods: Data collection, management, and analysis 55 56 57 58 59 60

3 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 37 of 39 BMJ Open

1 2 Data collection 18a Plans for assessment and collection of outcome, baseline, and other 3 methods trial data, including any related processes to promote data quality (eg, BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 duplicate measurements, training of assessors) and a description of 5 6 study instruments (eg, questionnaires, laboratory tests) along with 7 their reliability and validity, if known. Reference to where data 8 collection forms can be found, if not in the protocol 9 Page 8, lines 9-27 10 11 18b Plans to promote participant retention and complete follow-up, 12 13 including list of any outcome data to be collected for participants who 14 discontinue or deviate from intervention protocols 15 Not applicable as trial is nested within a Health and Demographic 16 Surveillance System 17 18 Data For19 Plans peer for data reviewentry, coding, security, only and storage, including any 19 management related processes to promote data quality (eg, double data entry; 20 21 range checks for data values). Reference to where details of data 22 management procedures can be found, if not in the protocol 23 Page 9, lines 15-22. 24 25 Statistical 20a Statistical methods for analysing primary and secondary outcomes. 26 methods Reference to where other details of the statistical analysis plan can be 27 28 found, if not in the protocol 29 Appendix 1 and appendix 2. 30 31 20b Methods for any additional analyses (eg, subgroup and adjusted 32 analyses) 33 Appendix 2 34 35 20c Definition of analysis population relating to protocol non-adherence 36 37 (eg, as randomised analysis), and any statistical methods to handle http://bmjopen.bmj.com/ 38 missing data (eg, multiple imputation) 39 Appendix 1 and appendix 2. 40 41 Methods: Monitoring 42 43 Data monitoring 21a Composition of data monitoring committee (DMC); summary of its role 44 and reporting structure; statement of whether it is independent from 45 on September 24, 2021 by guest. Protected copyright. 46 the sponsor and competing interests; and reference to where further 47 details about its charter can be found, if not in the protocol. 48 Alternatively, an explanation of why a DMC is not needed 49 Page 11, lines 3-14 50 51 21b Description of any interim analyses and stopping guidelines, including 52 who will have access to these interim results and make the final 53 54 decision to terminate the trial 55 No interim analyses will be conducted 56 57 58 59 60

4 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 38 of 39

1 2 Harms 22 Plans for collecting, assessing, reporting, and managing solicited and 3 spontaneously reported adverse events and other unintended effects BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 of trial interventions or trial conduct 5 6 This is part of the main outcome. 7 8 Auditing 23 Frequency and procedures for auditing trial conduct, if any, and 9 whether the process will be independent from investigators and the 10 sponsor 11 N/A 12 13 14 Ethics and dissemination 15 Research ethics 24 Plans for seeking research ethics committee/institutional review board 16 17 approval (REC/IRB) approval 18 ForPage peer 10 line 22review to page 11 line only2 19 20 Protocol 25 Plans for communicating important protocol modifications (eg, 21 amendments changes to eligibility criteria, outcomes, analyses) to relevant parties 22 (eg, investigators, REC/IRBs, trial participants, trial registries, journals, 23 24 regulators) 25 Page 11 line 20-22 26 27 Consent or assent 26a Who will obtain informed consent or assent from potential trial 28 participants or authorised surrogates, and how (see Item 32) 29 Page 7, lines 3-15 30 31 26b Additional consent provisions for collection and use of participant data 32 33 and biological specimens in ancillary studies, if applicable 34 Not applicable 35 36 Confidentiality 27 How personal information about potential and enrolled participants will 37 be collected, shared, and maintained in order to protect confidentiality http://bmjopen.bmj.com/ 38 before, during, and after the trial 39 40 All information is treated confidentially, and all publication and sharing 41 of data will not contain identifying information. 42 43 Declaration of 28 Financial and other competing interests for principal investigators for 44 interests the overall trial and each study site 45 Page 15, lines 12-13 on September 24, 2021 by guest. Protected copyright. 46 47 Access to data 29 Statement of who will have access to the final trial dataset, and 48 disclosure of contractual agreements that limit such access for 49 50 investigators 51 Page 15, lines 14-15 52 53 Ancillary and 30 Provisions, if any, for ancillary and post-trial care, and for 54 post-trial care compensation to those who suffer harm from trial participation 55 No plans, but participant insurance are in place. 56 57 58 59 60

5 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 39 of 39 BMJ Open

1 2 Dissemination 31a Plans for investigators and sponsor to communicate trial results to 3 policy participants, healthcare professionals, the public, and other relevant BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 groups (eg, via publication, reporting in results databases, or other 5 6 data sharing arrangements), including any publication restrictions 7 Page 11, lines 15-19 8 9 31b Authorship eligibility guidelines and any intended use of professional 10 writers 11 Not applicable 12 13 31c Plans, if any, for granting public access to the full protocol, participant- 14 level dataset, and statistical code 15 16 No plans 17 18 Appendices For peer review only 19 20 Informed consent 32 Model consent form and other related documentation given to 21 materials participants and authorised surrogates 22 Can be obtained upon contact to corresponding author 23 24 Biological 33 Plans for collection, laboratory evaluation, and storage of biological 25 26 specimens specimens for genetic or molecular analysis in the current trial and for 27 future use in ancillary studies, if applicable 28 No plans 29 30 *It is strongly recommended that this checklist be read in conjunction with the SPIRIT 2013 31 Explanation & Elaboration for important clarification on the items. Amendments to the 32 protocol should be tracked and dated. The SPIRIT checklist is copyrighted by the SPIRIT 33 34 Group under the Creative Commons “Attribution-NonCommercial-NoDerivs 3.0 Unported” 35 license. 36 37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 45 on September 24, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

6 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from

Can earlier BCG vaccination reduce early infant mortality? Study protocol for a cluster randomised trial in Guinea- Bissau ForJournal: peerBMJ Open review only Manuscript ID bmjopen-2018-025724.R2

Article Type: Protocol

Date Submitted by the 22-Aug-2019 Author:

Primary Subject Global health Heading:

Secondary Subject Heading: Epidemiology, Paediatrics http://bmjopen.bmj.com/ BCG vaccine, Oral polio vaccine, Cluster randomised trial, Non-specific Keywords: effects of vaccines

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 1 of 40 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Can earlier BCG vaccination reduce early infant mortality? Study protocol for a cluster 5 6 randomised trial in Guinea-Bissau 7 8 1,2,3,4 3,5 2 9 Sanne Marie Thysen , Aksel Karl Georg Jensen , Amabelia Rodrigues , Igualdino da Silva 10 Borges2, Peter Aaby2,3, Christine Stabell Benn1,2,3, Ane Bærent Fisker1,2,3 11 12 1 13 OPEN, Institute of Clinical Research, University of Southern Denmark, Odense, Denmark 14 2 Bandim Health Project, Bissau, Guinea-Bissau 15 16 3 Research Centre for Vitamins and Vaccines, Bandim Health Project, Statens Serum Institut, 17 18 Copenhagen, DenmarkFor peer review only 19 4 20 Center for Global Health (GloHAU), Aarhus University, Aarhus, Denmark 21 5 Section of Biostatistics, Department of Public Health, University of Copenhagen, Copenhagen, 22 23 Denmark 24 25 26 27 Corresponding author: Sanne Marie Thysen, OPEN, Institute of Clinical Research, University of 28 29 Southern Denmark, Odense, Denmark; [email protected] 30 31 32 33 34 35 36 37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 45 on September 24, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 2 of 40

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Abstract 5 6 7 Introduction 8 9 The Bacillus Calmette-Guérin vaccine (BCG) is designed to protect against tuberculosis, but the 10 11 vaccine may have broader effects. In 2014, the Strategic Advisory Group of Experts on 12 13 Immunization reviewed the literature on non-specific effects of BCG, and concluded that the 14 evidence was consistent with beneficial non-specific effects and requested further randomised trials. 15 16 17 Methods and analyses 18 For peer review only 19 Within the Bandim Health Project’s urban and rural Health and Demographic Surveillance Systems, 20 21 we will conduct a cluster-randomised trial in six suburban districts and 55 rural villages. Infants are 22 23 enrolled at a home visit before 72 hours of life. In intervention clusters, children are vaccinated with 24 BCG and oral polio vaccine (OPV). In control clusters, the caregivers are informed about 25 26 vaccination opportunities. Using Cox-proportional hazards models, we will test whether BCG and 27 28 OPV provided at a single home visit can reduce early infant mortality up to 60 days. 29 30 The trial was initiated with a pilot study in Biombo region in June 2015. The trial was scaled up to 31 32 full study including Oio and Cacheu regions in July 2016. The trial was expanded to include the 33 34 urban study area in July 2017. 35 36 Ethics and dissemination 37 http://bmjopen.bmj.com/ 38 BCG vaccination is delayed in many low-income settings. WHO-recommended home visits are 39 40 resource demanding and vaccines are not part of the recommendation. Utilizing the home visits to 41 42 provide BCG and OPV may provide countries with a further incentive to introduce a single home 43 44 visit. In countries, where home visits are already in place, vaccines can easily be added to reduce 45 early infant mortality. The trial is approved by the Guinean Ethical Committee (Reference number: on September 24, 2021 by guest. Protected copyright. 46 47 0016/CNES/INASA/2015) and the Danish Ethics Committee has given its consultative approval. 48 49 The results of the trial will be published in international peer-reviewed journals. 50 51 Trial registration 52 53 54 The trial is registered at clinicaltrials.gov (Clinicaltrials.gov ID: NCT02504203). 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 3 of 40 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 For peer review only 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 45 on September 24, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 4 of 40

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 5 of 40 BMJ Open

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 6 of 40

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Guinea-Bissau18. A major reason for this delay is the focus on reducing wastage of vaccines. 5 6 Freeze-dried vaccines like BCG have to be used within 6 hours after reconstitution with a diluent. 7 8 Therefore, a vial of BCG vaccine is often not opened unless 10-12 children are present for 9 10 vaccination. Hence, there are many missed vaccination opportunities. If BCG vaccine has profound 11 effects on neonatal mortality and morbidity many lives could be saved by providing BCG earlier. 12 13 14 In the present study, we aim to study the effect of BCG and OPV vaccinations provided at a single 15 home visit within 72 hours after birth on early infant mortality and morbidity among infants in a 16 17 rural and an urban setting in Guinea-Bissau. We hypothesise that BCG and OPV at birth provided at 18 For peer review only 19 a single home visit shortly after birth reduce early infant non-accidental mortality by 40%. 20 21 METHODS AND ANALYSES 22 23 Setting 24 25 The study is conducted in Bandim Health Project’s (BHP) Health and Demographic Surveillance 26 27 System (HDSS) sites in rural and urban Guinea-Bissau. 28 29 Rural site: The rural HDSS was established in 1990. The BHP teams survey women of fertile age 30 31 and their children below the age of 5 years in randomly selected clusters of villages in all the nine 32 health regions in the country. Children in the three rural regions closest to Bissau (Oio, Biombo and 33 34 Cacheu) are eligible for the study. 35 36 37 The rural clusters are followed with two-monthly visits by the mobile teams. At all visits, the http://bmjopen.bmj.com/ 38 women are asked whether they are pregnant. When a pregnancy is registered; the woman’s 39 40 nutritional status is assessed by measurement of a mid-upper-arm-circumference (MUAC); 41 42 information on antenatal care is collected prior to giving birth, as well as at the first visit after 43 delivery. Socio-economic factors (type of roofing, type of bathroom, possession of a mobile phone, 44 45 radio and generator) are registered. After the delivery, information on the place of delivery (home, on September 24, 2021 by guest. Protected copyright. 46 47 health facility) and who assisted the delivery is collected. 48 49 Urban site: The urban HDSS was established in 1978. Field assistants follow children below the 50 51 age of 3 years through home visits every third month. Pregnancies are registered every month and 52 53 followed monthly until birth. When a pregnancy is registered information on ethnic group, 54 gestational age, use of mosquito net is collected. After the delivery, information on place of delivery 55 56 (home, health facility) is collected. All children in the urban study area will be eligible for the study. 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 7 of 40 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Study nurses 5 6 Prior to study start, we selected study nurses to conduct home visits. In the rural study area, study 7 8 nurses were selected in collaboration with the regional health authorities. The nurses were selected 9 10 among nurses working at a health centre with a catchment area corresponding to the study clusters, 11 so the home visits can be done in addition to their normal routines. For the urban study area, the 12 13 BHP employed a study nurse fulltime. 14 15 Community Key Informants 16 17 In the rural areas, to ensure that the nurses are informed immediately about deliveries, we use 18 For peer review only 19 community key informants (CKI). CKIs were selected among residents in each rural village to 20 21 collect information about pregnancies, deliveries and deaths. The CKI communicate immediately 22 any delivery in the village to the study nurse. In the urban area, we take advantage of the close 23 24 follow-up by the field assistants, who circulate the zones daily. 25 26 Study design and randomisation 27 28 The trial is a cluster-randomised trial, randomising clusters (village clusters/sub-district zones) to 29 30 two different treatment groups, stratified by region and pre-trial mortality level (high/low). 31 32 Clusters, defined as village cluster in the rural area, and as sub-district zones in the urban area, were 33 34 randomised prior to study start using computer generated random numbers. All newborns are visited 35 36 as soon as possible after birth. In half of the clusters, we provide standard care and vaccines (BCG, 37 OPV) to children at the home visits; in the remaining clusters, children only receive standard care http://bmjopen.bmj.com/ 38 39 but no vaccines (Table 1). The trial is unblinded. 40 41 42 Inclusion criteria 43 All children registered during pregnancy are eligible for the study. Children can only be enrolled if 44 45 they are visited within 72 hours after birth. on September 24, 2021 by guest. Protected copyright. 46 47 Exclusion criteria 48 49 There are few exclusion criteria, because the study is expected to answer a pragmatic question about 50 51 the effect of BCG and OPV vaccination at home visits shortly after birth. 52 53  Children already BCG vaccinated 54 55  Moribund children (Expected not to survive the next 24 hours, as evaluated by the study nurse 56 57 at the enrolment visit) 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 8 of 40

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4  Children in rural villages where the BHP mobile teams coincidentally were in the village the 5 6 same day (and vaccinated all children) 7 8 9 Informed consent 10 11 All pregnant women receive an oral and a written explanation of the study, and are offered to 12 participate in the study at the time of registration of the pregnancy. In the rural area, the explanation 13 14 is given by the BHP nurse at the two-monthly visits. In the urban area, the explanation is given by 15 16 the fieldwork assistant at a monthly visit. It is explained that normally newborn children receive 17 their first vaccinations at the health centres. However, studies have suggested that it might be 18 For peer review only 19 beneficial to receive BCG and OPV early and the BHP will therefore test the effect of providing the 20 21 vaccines at home visits. All children receive a home visit with standard care. Children in half the 22 23 clusters (intervention clusters) receive BCG and OPV at enrolment. Children in the control clusters 24 can seek vaccination elsewhere and will be offered the BCG vaccine and OPV at the next home 25 26 visit by the BHP team if they did not get it elsewhere. If the pregnant woman accepts to take part in 27 28 the study, they are asked to sign or fingerprint a consent form. An independent witness confirms the 29 30 consent process by co-signing the form. 31 32 Enrolment 33 34 The study nurse visits every newborn child shortly after a CKI/mother calls, if possible on the same 35 day. At the home visit, the nurse asks the mother for confirmation of her consent for participation in 36 37 the study before revealing the randomisation. If the mother is not able to present the signed consent http://bmjopen.bmj.com/ 38 39 form, new consent forms are filled in. The nurse brings a questionnaire, a sticker with study number 40 41 and a vaccination card. 42 43 Intervention 44 45 For all children, the nurse examines and weighs the child, encourages skin-to-skin contact to keep on September 24, 2021 by guest. Protected copyright. 46 the newborn warm and if necessary performs umbilical cord care1. Following interview and 47 48 examination, group allocation is disclosed to the mother. For children in the control clusters, the 49 50 nurse informs about vaccination opportunities (vaccination at closest health centre or vaccination by 51 1 52 BHP nurse at next visit), as recommended by WHO . For children in the intervention clusters, the 53 nurse administers BCG and OPV (Table 1). 54 55 56 BCG vaccine 57 BCG vaccination is administered by intradermal injection; after vaccination most children develop a 58 59 scar at the injection site. Among BCG-vaccinated children, having a BCG scar is associated with 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 9 of 40 BMJ Open

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 10 of 40

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Outcomes 5 6 The primary outcome is non-accidental mortality rate between enrolment visit or 24 hours after 7 8 birth, whichever comes last, and 60 days of life. Follow-up time will be censored at the first of visit 9 10 by the BHP, death due to accident, date of registration of first non-trial vaccine given after 11 enrolment, or migration. All children living in intervention and control clusters are followed 12 13 through the HDSS routines with collection of information on vital status for all children. Deaths are 14 15 classified as accidental or non-accidental based on information from verbal autopsies. 16 17 Secondary outcomes and potential effect modifiers 18 For peer review only 19 We will evaluate the effect on severe morbidity considered as the composite outcome of non- 20 21 accidental death and non-accidental hospital admission. With declining mortality, we may not be 22 able to obtain a conclusive result for the primary outcome. We have defined the sample size for the 23 24 composite outcome, but retained mortality as the primary outcome. 25 26 We will conduct secondary analyses to assess whether an effect vary by low birthweight (<2500g), 27 sex, maternal BCG scarring, and season to gain a better knowledge of the potential effect modifiers 28 29 (a list of all outcomes can be found in Table 2). 30 31 We will study the cost-effectiveness of providing BCG and OPV at a home visit using the effects on 32 33 mortality and hospital admission from the present project. 34 35 For all deaths a verbal autopsy will be made. These will mainly be used to classify a death as 36 37 accidental or non-accidental, but where possible, we will also report cause-specific mortality. http://bmjopen.bmj.com/ 38 39 Data management 40 41 All data is collected on paper forms and entered in the BHP office. Records with registration of all 42 43 pregnant women are copied from the data tables of the routine HDSS to the BCG trial data table. 44 Forms for follow-up are pre-printed with identification information of the enrolled children. For 45 on September 24, 2021 by guest. Protected copyright. 46 routine HDSS visits, a list of all pregnant women is printed prior to the visit to ensure that all pregnant 47 48 women are invited to participate in the study prior to giving birth. Following data entry, the data is 49 50 checked for consistency. Inconsistencies between the BCG trial data and the HDSS data are checked. 51 Main outcome events are reviewed individually. 52 53 54 Statistical analyses 55 General analysis principles applied in all analyses can be found in appendix 1. In Cox-proportional 56 57 hazards models with age as underlying time-scale, we will assess the effect of providing BCG and 58 59 OPV at a single home visit. We will allow for different baseline hazards according to the factors 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 11 of 40 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 used in the randomisation (region and pre-study mortality level) and sex. Children enter the analysis 5 6 on the date of enrolment or 24 hours of age, whichever comes last, and remain in the analysis until 7 8 the first subsequent visit by the BHP team, date of registering first non-trial vaccine after enrolment, 9 10 death or migration within the first 60 days of life. In secondary analyses, we will investigate 11 whether the effect of BCG+OPV differ by the following potential effect modifiers, which in prior 12 13 trials have been important determinants of the effect: low birthweight (<2500 g), sex, maternal 14 15 BCG scarring, and season. Further details are provided in appendix 2. 16 17 Secondary outcomes are non-accidental hospital admissions, severe morbidity (composite outcome 18 For peer review only 19 of non-accidental mortality and non-accidental hospital admissions), consultations, growth (mid- 20 21 upper-arm circumference and weight-for-age z-score), and BCG scarring. Analyses are described in 22 appendix 2. We will furthermore perform a cost effectiveness analysis seeking to measure the cost 23 24 per death averted using a societal perspective, contrasting the costs of vaccine provision in the 25 26 present programme and an outreach system as tested in the trial. The final analysis plan is deposited 27 with the Data Safety and Monitoring Board. A list of all planned analyses is provided in appendix 2. 28 29 30 Sample size considerations 31 32 We estimated the design effect to 1.35 (ratio of square of the standard errors for the cluster- 33 adjusted/unadjusted HRs) based on pre-trial event data (deaths and hospitalisation) from the rural 34 35 HDSS clusters, where the study is conducted. We base our sample size calculation on the combined 36 37 secondary outcome for feasibility reasons, but retain mortality as the primary outcome. We http://bmjopen.bmj.com/ 38 estimated the proportion of events to be 2.4% in the absence of our intervention (prior data). Thus, 39 40 in order to obtain 80% power to detect a reduction in early infant severe morbidity if the true 41 42 reduction of BCG and OPV provided at home visits is larger than 40%, we will need to enrol at 43 44 least 6666 children. This sample size has 74% power to detect a significant difference in mortality if 45 the real effect is a 40% reduction from a mortality level of 2.1%. on September 24, 2021 by guest. Protected copyright. 46 47 48 Time schedule 49 50 The trial was initiated with a pilot study in Biombo region in June 2015. The trial was scaled up to 51 52 full study including Oio and Cacheu regions in July 2016. The trial was expanded to include the 53 urban study area in July 2017. In March 2019, the trial was stopped in the urban study area and in 54 55 the Cacheu region, due to lower than anticipated number of enrolments and number of events, 56 57 resulting in a lower than anticipated power. As a result of the lower than anticipated power, we plan 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 12 of 40

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 to end the trial prematurely in Oio and Biombo health regions in September 2019. The results of the 5 6 trial will be reported as planned despite the lower than anticipated power. 7 8 9 ETHICS AND DISSEMINATION 10 11 Ethical considerations 12 13 BCG is recommended at birth but vaccination is often delayed; in rural Guinea-Bissau most 14 18 15 children receive BCG after 1 month . While there is now increasing acceptance that BCG may 16 have beneficial non-specific effects, we are still a long way from altering the implementation of the 17 18 vaccination programme.For The proposal peer compares review two ways of deliveringonly BCG and OPV; both are an 19 20 improvement relative to the current situation in rural Guinea-Bissau, where less than 40% of all 21 25 22 infants get BCG during the first month of life . Hence, no child receives BCG later than it would 23 have done, had the trial not been carried out. The trial is approved by the Guinean Ethical 24 25 Committee (Reference number: 0016/CNES/INASA/2015) and the Danish Ethics Committee has 26 27 given its consultative approval. The trial is registered at clinicaltrials.gov (Clinicaltrials.gov ID: 28 NCT02504203). 29 30 31 Safety monitoring 32 33 A Data Safety and Monitoring Board has been formed consisting of a paediatrician (Anja Poulsen, 34 Rigshospitalet, Denmark), a statistician (Morten Frydenberg, Aarhus University, Denmark) and an 35 36 epidemiologist (Torben Sigsgaard, Aarhus University, Denmark). The members have been 37 http://bmjopen.bmj.com/ 38 appointed on their experience, reputation for objectivity, absence of conflicts of interest, and 39 knowledge on clinical trial methodology. 40 41 42 The vaccines used for the trial are prequalified by WHO to be given at birth. Adverse reactions are 43 44 rare for both BCG and OPV. At the two first visits after enrolment in the study, the BHP nurse 45 examines the BCG vaccination site and the axillary lymph glands of all children to assess on September 24, 2021 by guest. Protected copyright. 46 47 suppurative lymphadenitis as an adverse reaction to the BCG vaccination. Other serious adverse 48 49 events are captured through primary and/or secondary outcomes (mortality, hospitalisations and 50 51 consultations). 52 53 Dissemination 54 55 The results of the study will be published in international peer-reviewed journals and results will be 56 57 communicated to the national institute of public health in Guinea-Bissau. After publication of the 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 13 of 40 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 main results on completion of the trial, data requests can be submitted to the researchers at Bandim 5 6 Health Project. 7 8 9 Protocol amendments 10 11 Any protocol modifications will be discussed with the Data Safety and Monitoring Board, and 12 13 changes will be added to the trial registration. 14 15 Patient and public involvement 16 17 The communities were involved in locating households, when the HDSS was setup and contributed 18 For peer review only 19 information allowing tracing of internal migrants between villages throughout the study period. No 20 21 participants were involved in setting the research question or the outcome measure, nor were they 22 23 involved in developing plans for recruitment, design, or implementation of the study. Local 24 community key informants were selected to ensure timely information about birth. Nurses from 25 26 local health centres are involved in the study and have been trained to perform enrolment. This 27 28 ensures a close collaboration with the local health system. No participant was asked to advise on 29 interpretation or writing up the results. The results are disseminated to the national public health 30 31 institute. There are no plans to disseminate the results of the research to study participants. Study 32 33 results will be disseminated to local nurses involved in the study. 34 35 36 37 http://bmjopen.bmj.com/ 38 DISCUSSION 39 40 WHO recommends BCG and OPV at birth. However, due to the focus on not wasting vaccine 41 42 doses, BCG vaccination is often delayed18 26-28. Previous trials from Guinea-Bissau have showed 43 6 7 9 44 that early vaccination may have major impact on neonatal mortality among low-weight children , 45 and observational studies have suggested that BCG also has beneficial NSEs among normal-birth- on September 24, 2021 by guest. Protected copyright. 46 47 weight children3 11. In 2014, the Strategic Advisory Group of Experts on Immunization reviewed 48 49 the non-specific effects of the BCG vaccine, and concluded that the evidence was consistent with 50 beneficial NSEs. However, the bulk of evidence came from observational studies and they therefore 51 52 called for randomised trials. To our knowledge, this will be the first randomised trial assessing the 53 54 NSEs of BCG and OPV vaccination among children in a rural low-income setting. The SAGE 55 56 review did not include OPV, but a recent trial from Guinea-Bissau randomised children to BCG 57 only or BCG and OPV, and found that OPV was associated with a 32% reduction in mortality29. 58 59 Thus, both BCG and OPV may have beneficial NSEs. The results will therefore be very important 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 14 of 40

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 in broadening our understanding of the NSEs of BCG and OPV at birth. We will however not be 5 6 able to distinguish the effects of BCG and OPV in the present trial. 7 8 9 WHO recommends three home visits after birth, which is resource demanding and not implemented 10 in many low-income countries. It is not recommended that the home visits are utilised for early 11 12 vaccination. However, we hypothesise that a single home visit after birth will have major effects on 13 14 early infant mortality if BCG and OPV vaccination are added to the home visit. If BCG and OPV 15 provided at a single home visit can reduce early infant mortality, this should be included in the 16 17 WHO recommendations. A single home visit with vaccines will be easier to implement in a strained 18 For peer review only 19 health system than three home visits within 7 days. In countries, where home visits are already in 20 21 place, vaccines can easily be added to reduce early infant mortality. 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 45 on September 24, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 15 of 40 BMJ Open

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 16 of 40

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 17 of 40 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Authors’ contributions 5 6 AR, PA and CB conceived the idea for the trial. SMT, PA, CB and ABF designed the study. SMT 7 8 set up the study with help from ABF. SMT, IDSB and ABF supervised data collection. SMT, 9 10 AKGJ, and ABF planned the data analyses. SMT drafted the manuscript and data analysis plan with 11 help from AKGJ and ABF. All authors read and approved the final manuscript and data analysis 12 13 plan. 14 15 16 Acknowledgements 17 We thank our Data Safety and Monitoring board, Anja Poulsen, Morten Frydenberg, and Torben 18 For peer review only 19 Sigsgaard, for insightful comments on planned analysis, study design and implementation. 20 21 22 Funding 23 24 Karen Elise Jensen foundation is the main funder of the trial. Laerdals Foundation supports the trial. 25 The Research Centre for Vitamins and Vaccines, Bandim Health Project receives support from the 26 27 Danish National Research Foundation [DNRF108]. The funders had no role in the design of the 28 29 study or writing of the protocol. 30 31 32 Competing interests 33 The authors declare no competing interests. 34 35 Access to data 36 37 Data is available on a collaborative basis, contact [email protected]. http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 45 on September 24, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from

BMJ Open Page 18 of 40

1 2 3 4 Figure 1: Recommended vaccination schedule in Guinea-Bissau 5 6 7 8 9 10 11 12 For peer review only 13 14 15 16

17 http://bmjopen.bmj.com/ 18 19 20 21 22 23 24

Page 19 of 40 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 Table 2: Outcomes 5 6 Primary outcome 7 1 8  Early infant non-accidental mortality 9 Potential effect modifiers 10  Low birthweight defined as birthweight lower than 11 2500g 12  Sex 13 14  Maternal BCG scarring 15  Season 16 Secondary outcomes 17  Non-accidental hospital admission1 18 For peer review only 19  Severe morbidity (composite outcome of non-accidental 1 20 mortality and non-accidental hospital admissions) 21  Consultations2 22  Growth3 23  Mid-upper-arm circumference 24 25  Weight-for-age z-score 26  BCG scarring2 27  Cost-effectiveness of providing BCG and OPV at home 28 visits 29  Cause-specific mortality 30 1 31 Analysed using Cox proportional hazards models. Further details are provided in Appendix 2. 32 2 Analysed using Log-binomial regression. Further details are provided in Appendix 2. 33 3 Analysed using Linear regression. Further details are provided in Appendix 2. 34 35 36 37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 45 on September 24, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 21 of 40 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For peer review only 17 18 19 20 21 22 Figure 1: Recommended vaccination schedule in Guinea-Bissau 23 24 25 26 27 28 29 30 31 32

33 http://bmjopen.bmj.com/ 34 35 36 37 38 39 40

1 2

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 23 of 40 BMJ Open

1 2

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 24 of 40

1 2

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 25 of 40 BMJ Open

1 2

18 For peer- 60 days review only 19 - Death due to accident 20 - Date of registering first non-trial vaccine given after 21 22 enrolment 23 - Migration (migration out of the study area as per HDSS 24 definition) 25 26 Time scale Age 27 Failure definition Death 28 29 Statistical tool Cox proportional hazards model 30 Stratification We will employ models that allow for different baseline hazards 31 according to sex, region and pre-study mortality level 32 33 Clustering We will use robust standard errors to account for clustering within 34 village cluster 35 Outline stata code 36 37 For analysis: stset outdate, f(dead==1) enter (max(dob+1, date_enrol) /// http://bmjopen.bmj.com/ 38 exit (min ( dnasc+60, date_mobileteam) origin(dob) 39 40 41 stcox random, strata(sex reg prmorlev) vce(cl regam) 42 43 For model checking: estat phtest, detail† 44 45 stphplot, strata(random) adj(sex reg prmorlev) on September 24, 2021 by guest. Protected copyright. 46 stcox random, strata(sex reg prmorlev) vce(cl regam) tvc(random) 47 texp(_t) 48 49 50 *reg=region, prmorlev=pre-study mortality level, regam=village 51 cluster 52 † If the proportional hazards assumption is not fulfilled we will still report the marginal hazard ratio 53 54 but it will be supplemented by hazard ratios for two-three categorized time periods. 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 26 of 40

1 2

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 27 of 40 BMJ Open

1 2

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 28 of 40

1 2

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 29 of 40 BMJ Open

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2018-025724 on 24 September 2019. Downloaded from 4 - 60 days 5 6 - Migration (migration out of the study area as per HDSS 7 definition) 8 Time scale Age 9 10 Failure definition First hospital admission – only overnight hospitalisations or arrival at 11 the hospital and death within the first day will be considered in this 12 13 analysis. 14 Statistical tool Cox proportional hazards model 15 16 Stratification We will employ models that allow for different baseline hazards 17 according to sex, region and pre-study mortality level 18 Clustering ForWe peer will use robust review standard errors only to account for clustering within 19 20 village cluster 21 Outline stata code 22 For analysis: stset outdate, f(hosp==1) enter (max(dob+1, date_enrol) /// 23 24 exit (min ( dnasc+60, date_mobileteam) origin(dob) 25 26 stcox random, strata(sex reg prmorlev) vce(cl regam) 27 28

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 30 of 40

1 2

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 31 of 40 BMJ Open

1 2

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 32 of 40

1 2

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 33 of 40 BMJ Open

1 2

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 34 of 40

1 2

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 35 of 40 BMJ Open

1 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 36 of 40

2 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 37 of 40 BMJ Open

3 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 38 of 40

4 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 39 of 40 BMJ Open

5 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 40 of 40

6 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml