Maternal immunisation with trivalent inactivated influenza vaccine for prevention of influenza in infants in Mali: a prospective, active-controlled, observer-blind, randomised phase 4 trial

Summary Background Despite the heightened risk of serious influenza during infancy, vaccination is not recommended in infants younger than 6 months. We aimed to assess the safety, immunogenicity, and efficacy of maternal immunisation with trivalent inactivated influenza vaccine for protection of infants against a first episode of laboratory-confirmed influenza. Methods We did this prospective, active-controlled, observer-blind, randomised phase 4 trial at six referral centres and community health centres in Bamako, Mali. Third-trimester pregnant women (≥28 weeks' gestation) were randomly assigned (1:1), via a computer-generated, centre-specific list with alternate block sizes of six or 12, to receive either trivalent inactivated influenza vaccine or quadrivalent meningococcal vaccine. Study personnel administering vaccines were not masked to treatment allocation, but allocation was concealed from clinicians, laboratory personnel, and participants. Infants were visited weekly until age 6 months to detect influenza-like illness; laboratory-confirmed influenza diagnosed with RT-PCR. We assessed two coprimary objectives: vaccine efficacy against laboratory-confirmed influenza in infants born to women immunised any time prepartum (intention-to-treat population), and vaccine efficacy in infants born to women immunised at least 14 days prepartum (per-protocol population). The primary outcome was the occurrence of a first case of laboratory-confirmed influenza by age 6 months. This trial is registered with ClinicalTrials.gov, number NCT01430689. Findings We did this trial from Sept 12, 2011, to Jan 28, 2014. Between Sept 12, 2011, and April 18, 2013, we randomly assigned 4193 women to receive trivalent inactivated influenza vaccine (n=2108) or quadrivalent meningococcal vaccine (n=2085). There were 4105 livebirths; 1797 (87%) of 2064 infants in the trivalent inactivated influenza vaccine group and 1793 (88%) of 2041 infants in the quadrivalent meningococcal vaccine group were followed up until age 6 months. We recorded 5279 influenza-like illness episodes in 2789 (68%) infants, of which 131 (2%) episodes were laboratory-confirmed influenza. 129 (98%) cases of laboratory-confirmed influenza were first episodes (n=77 in the quadrivalent meningococcal vaccine group vs n=52 in the trivalent inactivated influenza vaccine group). In the intention-to-treat population, overall infant vaccine efficacy was 33·1% (95% CI 3·7–53·9); in the per-protocol population, vaccine efficacy was 37·3% (7·6–57·8). Vaccine efficacy remained robust during the first 4 months of follow-up (67·9% [95% CI 35·1–85·3] by intention to treat and 70·2% [35·7–87·6] by per protocol), before diminishing during the fifth month (57·3% [30·6–74·4] and 60·7 [33·8–77·5], respectively). Adverse event rates in women and infants were similar among groups. Pain at the injection site was more common in women given quadrivalent meningococcal vaccine than in those given trivalent inactivated influenza vaccine (n=253 vs n=132; p<0·0001), although 354 [92%] reactions were mild. Obstetrical and non-obstetrical serious adverse events were reported in 60 (3%) women in the quadrivalent meningococcal vaccine group and 61 (3%) women in the trivalent inactivated influenza vaccine group. Presumed neonatal infection was more common in infants in the trivalent inactivated influenza vaccine group than in those in the quadrivalent meningococcal vaccine group (n=60 vs n=37; p=0·02). No serious adverse events were related to vaccination. Interpretation Vaccination of pregnant women with trivalent inactivated influenza vaccine in Mali—a poorly resourced country with high infant mortality—was technically and logistically feasible and protected infants from laboratory-confirmed influenza for 4 months. With adequate financing to procure the vaccine, implementation will parallel the access to antenatal care and immunisation coverage of pregnant women with tetanus toxoid. Funding Bill & Melinda Gates Foundation.


Introduction
Pregnant women and young infants are at increased risk of developing severe, complicated, and sometimes fatal infl uenza infection; [1][2][3][4] however, no infl uenza vaccines are approved for infants younger than 6 months. [5][6][7] Maternal immunisation against infl uenza is a promising strategy to reduce disease in pregnant women and young infants. 8,9 Trials in Bangladesh 10 and South Africa 11 showed protection against laboratory-confi rmed infl uenza in infants born to mothers who received trivalent inactivated infl uenza vaccine, but additional health benefi ts in those infants (eg, higher birthweight and reduced likelihood of being small for gestational age) have been inconsistent. 11,12 Remaining questions include more precise determination of the duration of protection for infants that can accrue from maternal immunisation, 8 and the technical and logistical feasibility and eff ectiveness of implementation of programmes in resource-limited settings with high to moderate infant mortality rates. 13 We aimed to address these questions in the course of a post-licensure clinical trial of the safety, immunogenicity, and effi cacy of maternal infl uenza immunisation for prevention of infl uenza in infants younger than 6 months in Mali, west Africa-one of the world's least developed countries, with the world's seventh highest infant mortality rate. 14, 15 Mali, nevertheless, maintains a vaunted Expanded Program on Immunization (EPI) that includes immunisation of pregnant women with tetanus toxoid and the introduction of fi ve new EPI vaccines for infants since 2005. [16][17][18] Introduction of additional vaccines for pregnant women, particularly trivalent inactivated infl uenza vaccine, the composition of which changes annually, would be challenging, but would make use of an existing eff ective vaccine delivery platform.

Study design and participants
We did this prospective, active-controlled, observer-blind, randomised phase 4 trial at six referral centres and community health centres in Bamako, Mali. In the year before starting the trial, infl uenza activity occurred from September to April, with peaks in October and February.
Third-trimester pregnant women (≥28 weeks' gestation based on last menstrual period, ultrasound, or uterine height) presenting to participating health centres for prenatal care were eligible for inclusion. Participants had to be able to understand and comply with planned study procedures, provide written informed consent before initiation of any study procedures, and intend to reside in the study area until their newborn infants were 6 months old. Participants could not be members of a household that already had a woman who was participating or had

Research in context
Evidence before this study Immunisation of pregnant women against infl uenza has been common practice in industrialised nations since their vulnerability to severe disease and adverse outcomes was recognised. Nevertheless, this practice has not been adopted by low-income countries with constrained resources. When we undertook this trial, maternal immunisation as a strategy to prevent infant illness and avert associated morbidity and mortality was gaining traction. We searched PubMed between April 17, 1996, andMarch 8, 2016, for clinical trials with the terms "maternal infl uenza vaccination" and "maternal infl uenza immunization". Our search yielded 36 publications. Two publications were of randomised clinical trials done in low-income and middle-income countries that measured the effi cacy of maternal infl uenza vaccination in protection of infants. The fi rst trial, done in Bangladesh, reported 63% effi cacy in the reduction of laboratory-confi rmed infl uenza in infants aged up to 24 weeks. Furthermore, infants born to women who received infl uenza vaccine were less likely to be small for gestational age and had a higher mean birthweight than did those born to women in the control group. The next trial was done in South Africa and reported 48·8% effi cacy in infants aged up to 24 weeks; however, the other benefi ts to infants were not shown.

Added value of this study
Our trial represents the largest evaluation so far of maternal infl uenza vaccination as a strategy to prevent infl uenza in the youngest infants. Additionally, it is the fi rst such study to be completed in west Africa, specifi cally Mali, one of the poorest countries in the world. Demonstrating the effi cacy of maternal infl uenza vaccination in this setting is compelling. Moreover, establishing that effi cacy is highest in the fi rst 4 months of life (67·9%) is important as the duration of protection conferred through maternal vaccination and anticipated benefi ts are assessed. Finally, the absence of an eff ect on the incidence of low birthweight is consistent with fi ndings shown in South Africa.

Implications of all the available evidence
Our study unequivocally demonstrates effi cacy of maternal infl uenza vaccination against laboratory-confi rmed infl uenza in infants and mothers. Moreover, there was high acceptability and logistical feasibility. However, our trial and that done in South Africa did not corroborate the previously reported benefi ts on neonatal outcomes. Moreover, because these trials were not designed to measure an eff ect on severe, deadly disease, there remains a notable gap when assessing the cost-eff ectiveness of this intervention. Although the success of maternal tetanus immunisation programmes suggests that implementation of infl uenza vaccination would also be successful, the related cost would need to be justifi ed by the gains of the benefi ts aff orded so that local policy makers and donors could invest their restricted funds. participated in this study. Other exclusion criteria were a history of severe reactions following previous immunisation with infl uenza or meningococcal vaccines; Guillain-Barré syndrome; known allergy or hypersensitivity to eggs, egg proteins, latex, diphtheria toxoid, or any other components of trivalent inactivated infl uenza vaccine (Vaxigrip) and quadrivalent meningococcal conjugate vaccine (Menactra); known chronic medical disorder that, in the judgment of the investigator, could compromise assessment of the study vaccine or put the participant at risk; known active infection with HIV, hepatitis B virus, or hepatitis C virus; complications with the ongoing pregnancy, including preterm labour (with cervical change), placental abruption, premature rupture of membranes, known major congenital anomaly, or pre-eclampsia; acute illness or an oral temperature greater than or equal to 37·8°C within 72 h of vaccination (resulted in a temporary delay of vaccination); receipt of any other vaccine, excluding tetanus toxoid, within 2 weeks (for inactivated vaccines) or 4 weeks (for live vaccines and meningococcal A conjugate vaccine) before vaccination in this study; receipt of immunoglobulins or any blood products within 30 days before administration of study vaccines; chronic administration of immuno suppressants or other immune-modifying drugs within 90 days before administration of study vaccines; or any disorder that, in the opinion of the investigator, might compromise the wellbeing of the participant or compliance with study procedures, or interfere with the assessment of study vaccines. We additionally excluded women who intended to travel out of the study area in the 40 days after delivery. Enrolment continued until the requisite number of laboratory-confi rmed infl uenza cases was detected in infants born to vaccinated women.
Approval for the research was obtained from the University of Maryland, Baltimore Institutional Review Board; the ethics committee of the Faculté de Médecine, Pharmacie et Odonto-Stomatologie of Mali; and the Ministry of Health of Mali. Community sensitisation was achieved through community leaders, health centre representatives and community members who attended community-wide meetings. All participants provided informed consent. If the participant was illiterate, consent was obtained in the presence of a literate witness after listening to the audiotaped version of the consent form in Bambara, the local language.

Randomisation and masking
Participants were randomly allocated (1:1), via a computergenerated, centre-specifi c list with alternate block sizes of six or 12, to receive trivalent inactivated infl uenza vaccine (Vaxigrip, Sanofi Pasteur, Lyon, France) or quadrivalent meningococcal conjugate vaccine (Menactra, Sanofi Pasteur, Lyon, France). At enrolment, consenting participants were assigned an identifi cation number, which at vaccination was referenced on the randomisation list and the allocated treatment given. The identifi cation numbers for ineligible participants or those who withdrew before vaccination were not reassigned.
Study personnel who administered study vaccines and were aware of treatment allocation had no contact with participants after vaccination and were instructed not to reveal the identity of the study vaccines either to participants or to personnel masked to treatment allocation. Although the syringes used to administer the vaccines were diff erent in appearance, participants were instructed to look away from the vaccinator and were unaware of the assigned intervention.

Procedures
Quadrivalent meningococcal conjugate vaccine, rather than placebo, was given to controls to provide a potential benefi t for all participants in this poor, mostly illiterate, vulnerable population of pregnant women. Moreover, that vaccine was unlikely to interfere with the primary outcome of the trial, yet would provide protection against meningococcal disease. Although disease due to serogroup A has largely disappeared from the region, other serogroups continue to cause epidemics in Mali.
The composition of trivalent inactivated infl uenza vaccine, supplied in prefi lled syringes, changed during the trial. polysaccharides conjugated to diphtheria toxoid protein, was supplied in single-dose vials. A single 0·5 mL dose of trivalent inactivated infl uenza vaccine or quadrivalent meningococcal conjugate vaccine was injected into the deltoid muscle. Study vaccines were stored in secure, temperature-monitored refrigerators or cold rooms at 2-8°C.
After vaccination, women were observed for 30 min. 7 days after vaccination, fi eld personnel interviewed the women about any local and systemic reactions. 28 days after vaccination, participants were clinically evaluated. Additional visits to evaluate safety and immunogenicity in women and infants were done at delivery and when the infant was 3 months and 6 months old. Each evaluation included a physical examination and blood specimen collection. When available, the infant birth sample was cord blood; otherwise, the birth sample was collected within 7 days after birth. To determine gestational age at birth, the New Ballard Score was measured at delivery or within 7 days after birth. 19 Serious adverse events were recorded throughout study participation.
Besides safety follow-up visits, from enrolment to when the infant reached age 6 months, fi eld personnel undertook weekly visits to detect infl uenza-like illness and severe acute respiratory infection. During each visit, the participating woman and infant (if already born) had their temperatures measured and were examined for infl uenza-like illness; women were additionally examined for severe acute respiratory infection. When case defi nitions for either disease were met (appendix p 5), nasopharyngeal and oropharyngeal swabs and a malaria blood smear were obtained. If infl uenza was detected by RT-PCR, the case was deemed to be laboratory-confi rmed infl uenza. Standard-of-care treatment was off ered.
Because the primary objective was to measure the effi cacy of maternal immunisation for prevention of laboratory-confi rmed infl uenza in their infants younger than 6 months, women were withdrawn from weekly surveillance of infl uenza-like illness following stillbirth, fetal death, infant death, or other events that precluded infant surveillance. Nevertheless, safety follow-up of women continued until 6 months after delivery. Appendix p 6 describes methods for sample collection, RT-PCR to detect infl uenza virus, virus subtyping, and haemagglutination inhibition antibody measurement.

Outcomes
We assessed two coprimary objectives: vaccine effi cacy in infants born to women vaccinated any time prepartum (intention-to-treat analysis), and vaccine effi cacy in infants born to women vaccinated at least 14 days prepartum (perprotocol analysis). The primary outcome was the occurrence of a fi rst case of laboratory-confi rmed infl uenza by age 6 months. Secondary outcomes were the occurrence of a fi rst case of laboratory-confi rmed infl uenza in women (prepartum and post partum); occurrence of a fi rst case of febrile infl uenza-like illness by age 6 months in infants; occurrence of a fi rst case of febrile infl uenza-like illness in women (prepartum and post partum); occurrence of local and systemic reactogenicity after injection, related serious adverse events for the entire follow-up period, and all pregnancy complications; levels of infl uenza virus antibodies by haemagglutination inhibition before and 4 weeks after vaccination, at delivery, and 3 and 6 months after delivery. Tertiary outcomes included the frequency of each infl uenza virus type circulating in the study population, the levels of maternally derived infl uenza virus haemagglutination inhibition antibodies present in infants at birth and at ages 3 and 6 months, birthweights of infants born at a health centre, and the occurrence of severe acute respiratory infection in pregnant women. Appendix p 6 lists additional outcomes not included in the manuscript.

Statistical analysis
We calculated vaccine effi cacy with the formula: where VE is vaccine effi cacy, h is the ratio of follow-up time up to age 6 months in infants born to recipients of quadrivalent meningococcal vaccine to the follow-up time in infants born to recipients of trivalent inactivated infl uenza vaccine, and P is the proportion of all cases of laboratory-confi rmed infl uenza occurring by age 6 months in infants whose mothers received trivalent inactivated infl uenza vaccine. This calculation is equivalent to estimating vaccine effi cacy as 1-R, where R is the ratio of laboratory-confi rmed infl uenza incidence rates. We used the ratio of incidence rates, rather than the ratio of proportions of participants who had laboratory-confi rmed infl uenza to account for infants lost to follow-up before age 6 months. We estimated vaccine effi cacy in both the intention-to-treat and the perprotocol populations. Only infants' fi rst laboratoryconfi rmed infl uenza episodes were counted. Follow-up time was time from birth to fi rst case of laboratoryconfi rmed infl uenza, infants reaching age 6 months, or exiting the study. We calculated vaccine effi cacy for each month of age (0-5 months) and cumulative to each month of age.
For safety outcomes, we used Fisher's exact tests and Student's t tests to compare the proportion of participants who had each event per vaccine group. We did time-toevent analysis using Cox proportional hazards regression with laboratory-confi rmed infl uenza as the outcome to establish whether year of vaccination or timing of vaccination relative to delivery had an eff ect on effi cacy. Birthweight analysis was limited to weights that were either 500 g and more or 5000 g and less. We compared birthweight between vaccine groups both overall and within infl uenza seasons, defi ned as months with higherthan-average rates of laboratory-confi rmed illness (February to April, September to October).
Sample-size calculations were based on a comparison of the expected proportion, P, of all cases of laboratoryconfi rmed infl uenza that occurred by age 6 months in infants whose mothers received trivalent inactivated infl uenza vaccine to the null value, P 0 , using exact binomial calculations and assuming equal total followup time in each vaccine group (h=1). For the intentionto-treat analysis, we assumed a laboratory-confi rmed infl uenza attack rate of 2·2% by age 6 months in infants born to recipients of quadrivalent meningococcal vaccine and a 55% reduction in the attack rate in infants of recipients of trivalent inactivated infl uenza vaccine, to 0·99%; therefore, p=0·31034 and P 0 =0·5. For a onesided α of 0·025, 77 cases of laboratory-confi rmed infl uenza were needed to ensure 90% power for the intention-to-treat analysis, implying a need for about 4828 participants. Allowing for a 10% loss to follow-up, the sample size calculated became about 5370 participants. For the per-protocol analysis, we assumed vaccine effi cacy to be 60%-ie, a laboratoryconfi rmed infl uenza attack rate of 0·88% by age 6 months in infants born to recipients of trivalent See Online for appendix inactivated infl uenza vaccine. To ensure 90% power to show a vaccine effi cacy of more than 5%, 67 cases of laboratory-confi rmed infl uenza were needed, implying a sample size of 4352 participants. Allowing for a 20% loss to follow-up, or for the mother receiving vaccine less than 14 days before delivery, the sample-size requirement became about 5440 participants.
Enrolment was closed once 77 cases of infant laboratory-confi rmed infl uenza were recorded, but surveillance continued until the infants reached 6 months of age. A Data Safety Monitoring Board oversaw the study and reviewed data on a regular basis. We did analyses with Stata (version 14.0) and NCSS (version 10). We did power calculations with PASS  (version 12). This trial is registered with ClinicalTrials. gov, number NCT01430689.

Role of the funding source
The funder of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had fi nal responsibility for the decision to submit for publication.

Results
We did this trial from Sept 12, 2011, to Jan 28, 2014. Between Sept 12, 2011, and April 18, 2013, we randomly assigned 4193 women to receive trivalent inactivated infl uenza vaccine (n=2108) or quadrivalent meningococcal vaccine (n=2085; fi gure 1). Baseline characteristics were similar between groups (table 1). One (<1%) woman, who was inadvertently vaccinated twice (once with each vaccine), was followed up as part of her initial assignment group. 4087 (97%) women remained in the study until delivery; 3661 (87%) women were followed up until 6 months after delivery (fi gure 1). There were 4105 livebirths; 1797 (87%) of 2064 infants in the trivalent inactivated infl uenza vaccine group and 1793 (88%) of 2041 infants in the quadrivalent meningococcal vaccine group were followed up until age 6 months (fi gure 1). Due to political upheaval in Mali, study personnel were unable to do household visits for 1 week in March, 2012, and 2 weeks in May, 2012. We recorded 5279 infl uenza-like illness episodes in 2789 infants younger than 6 months, of which 131 (2%) episodes were laboratory-confi rmed infl uenza. 129 (98%) cases of laboratory-confi rmed infl uenza were fi rst episodes (n=77 in the quadrivalent meningococcal vaccine  *Incidence per 1000 infant-days of follow-up.

Table 2: Maternal infl uenza vaccine effi cacy against fi rst episodes of laboratory-confi rmed infl uenza in infants younger than 6 months born to women vaccinated at any time prepartum or 14 or more days prepartum
group vs n=52 in the trivalent inactivated infl uenza vaccine group). 116 (90%) fi rst episodes of laboratory-confi rmed infl uenza were in infants of women vaccinated at least 14 days prepartum. The 77 cases needed to complete vaccine effi cacy analyses were reached by April 16, 2013; surveillance of post-partum women and their infants continued until infants reached age 6 months. One episode of laboratory-confi rmed infl uenza was associated with malaria parasitaemia. Overall infant vaccine effi cacy was 33·1% (95% CI 3·7-53·9) in the intention-to-treat population, and 37·3% (7·6-57·8) in the per-protocol population (table 2). Notably, cumulative vaccine effi cacy in infants in the intention-totreat population was 67·9% in the fi rst 4 months of followup, fell to 57·3% at the fi fth month of surveillance, and dropped precipitously in the last month of follow-up, by which time protection was no longer evident (table 2). Cumulative vaccine effi cacy in infants in the per-protocol population was 70·2% in the fi rst 4 months of follow-up and 60·7% at the fi fth month of surveillance (table 2). Within the trivalent inactivated infl uenza vaccine group, Cox regression analysis of the relative risk of laboratoryconfi rmed infl uenza showed that risk of infl uenza decreased when trivalent inactivated infl uenza vaccine had been given at least 15 days prepartum (p=0·02; appendix p 7). As long as the vaccine was given at least 15 days before delivery, no additional benefi t was noted in women who had even longer intervals; women in neither the trivalent inactivated infl uenza vaccine group (Cox regression p=0·90) nor the quadrivalent meningococcal vaccine group (p=0·73) had a signifi cant change in rates of laboratory-confi rmed infl uenza as the time from delivery to vaccination increased above 14 days.
We measured haemagglutination inhibition antibody titres against infl uenza A/California/07/09 in 180 mother-infant pairs (fi gure 2). A subset of 43 pairs (plus one twin) constituted a nested case-control study in  which we tested samples from 11 H1N1 cases (including a pair of twins) and 33 birthdate-matched controls (plus or minus 30 days). The remaining 137 pairs included 46 pairs with an infant with laboratory-confi rmed illness, and 91 pairs with an infant who did not have laboratoryconfi rmed illness but might have had infl uenza-like illness; these pairs represented a convenience sample of participants who completed the study. By age 3 months, infant geometric mean titres (GMT) had decreased by more than 50%, although more infants in the trivalent inactivated infl uenza vaccine group had haemagglutination inhibition antibody titres of 40 or more (fi gure 2, appendix pp 10, 11). At age 6 months, haemagglutination inhibition antibody titres of 40 or more did not diff er signifi cantly between infants in either vaccine group, although maternal titres remained higher in the trivalent inactivated infl uenza vaccine group (appendix p 11). Of note, GMT increased in the quadrivalent meningococcal vaccine group at age 6 months, probably due to natural immunity acquired between ages 3 and 6 months. As haemagglutination inhibition antibody titres decreased with age in the trivalent inactivated infl uenza vaccine group, effi cacy also decreased (fi gure 2). The most frequently reported local and systemic reactions were pain at the injection site and febrile sensation (appendix p 12). Pain at the injection site was more commonly reported in women given quadrivalent meningococcal vaccine than in those given trivalent inactivated infl uenza vaccine (n=253 vs n=132; p<0·0001), although reactions were mostly mild (92%; appendix p 12). Rates of unrelated obstetrical and non-obstetrical serious adverse events in women were similar between groups (appendix p 13). The most commonly reported events were hypertensive disorders of pregnancy, which were equally common among both vaccine groups; 1% of participants in each group had pre-eclampsia (p=0·89; appendix p 13). No serious adverse event was related to study treatment. There were fi ve (<1%) unrelated deaths among study participants (n=2 in the trivalent inactivated infl uenza vaccine group and n=3 in the quadrivalent meningococcal vaccine group); two (40%) patients died due to obstetrical complications and three (60%) patients died after the 42 day period after delivery.
Although rates of serious adverse events in infants were similar between groups, presumed neonatal infection was more common in infants in the trivalent inactivated infl uenza vaccine group than in those in the quadrivalent meningococcal vaccine group (n=60 vs n=37; p=0·02; appendix p 14). No serious adverse events in infants were related to maternal vaccination. 89 infants died: 52 (59%) infants in the trivalent inactivated infl uenza vaccine group and 37 (41%) infants in the quadrivalent meningococcal vaccine group (p=0·13; appendix p 14); no deaths were due to laboratoryconfi rmed infl uenza. Appendix p 15 summarises the timing and causes of death.
Per-protocol analysis of the number of infants with a Ballard score less than 33 yielded an overall prematurity rate of 1·8%, which did not correlate with rates measured using date of last menstrual period or results of fi rsttrimester ultrasounds (appendix p 16). 358 (9%) liveborn infants were born at a low birthweight; there was no diff erence in birthweight between vaccine groups (p=0·20). Furthermore, there was no diff erence in birthweight among infants born during infl uenza season.

Discussion
Here we report results of the largest randomised controlled trial so far of trivalent inactivated infl uenza vaccine in pregnant women, which was undertaken in Mali, where infant and maternal mortality rates are among the world's highest. 14 Trivalent inactivated infl uenza vaccine elicited robust antibody responses and women and their infants were signifi cantly protected against laboratory-confi rmed infl uenza, corroborating results from Bangladesh (63% vaccine effi cacy, 95% CI 5-85) and South Africa (48·8%, 11·6-70·4), 10,11 and supporting WHO recommendations that pregnant women should be the highest priority target for infl uenza vaccination. 20 Because pregnant women and infants are at high risk for severe and fatal infl uenza illness even in affl uent countries, our fi ndings showing effi cacy of maternal immunisation in severely resource-constrained Mali, one of the world's least developed countries, constitute encouraging new information.
Maternal immunisation with trivalent inactivated infl uenza vaccine provided robust protection to infants during the fi rst 4 months of life. Thereafter, as haemagglutination inhibition antibody titres diminished, effi cacy decreased and was no longer evident at month 6 of follow-up. These observations support the contention that transplacental maternal antibody protects infants against   laboratory-confi rmed infl uenza. Haemagglutination inhibition antibody kinetics resembled those reported in Bangladesh 21 and South Africa, 22 and align with fi ndings from seroepidemiological studies 23,24 showing that by age 6 months most Malian infants no longer have protective titres of maternally derived measles and Haemophilus infl uenzae type b antibodies.
Infl uenza vaccine was well tolerated by pregnant women in our study, corroborating increasing evidence supporting the safety of trivalent inactivated infl uenza vaccine during pregnancy. 25 The Bangladesh trial reported that infants born during the infl uenza season to women who received infl uenza vaccine had higher birthweights than did those born to women who received control vaccine during that period. 12 In Mali and South Africa there was no benefi cial eff ect of maternal immunisation with trivalent inactivated infl uenza vaccine on birthweight in infants born anytime during the study, 11 including during infl uenza season. Exclusion of women with high-risk pregnancies from our study and inclusion of women late in pregnancy might have made it diffi cult to detect diff erences in birthweight due to maternal disorders. Moreover, infants born to women who had been in the study longer had higher birthweights than did those vaccinated closer to delivery, further decreasing the likelihood of detecting a diff erence between vaccine groups.
While demonstrating the effi cacy and safety of maternal infl uenza immunisation, we were also able to address the technical and logistical feasibility of implementation of such a programme in Mali. The trial was well received by the community as the study team worked at local health centres alongside routine prenatal care (that included the administration of tetanus toxoid) to enrol more than 4000 women. The workfl ow pattern for administration of study vaccine paralleled that of tetanus toxoid administration and was well accepted by local providers. Nevertheless, remaining aspects, such as the availability of an appropriate vaccine, access to prenatal care, and cost, would aff ect implementation of a maternal infl uenza immunisation programme.
We noted seasonal infl uenza peaks with diff erent infl uenza viruses from year to year, and these fl uctuations aff ected vaccine effi cacy. Notably, vaccinated mothers were signifi cantly protected against infl uenza B, whereas infants were not. The probable explanation relates to when diff erent B-virus lineages circulated. Cases of type B laboratory-confi rmed infl uenza in infants born to mothers vaccinated with trivalent inactivated infl uenza vaccine were Yamagata lineage infections, whereas mothers had received B/Brisbane/60/2008 (Victoria lineage) vaccine. By the time infants born to women who received Yamagata lineage-containing vaccine (B/Wisconsin/1/2010) were exposed to infl uenza, little type B was circulating. This fi nding shows the complexity of vaccine selection and supports the use of quadrivalent infl uenza vaccines containing both type-B lineages.
Timeliness of importation of newly formulated vaccine, promptness of initiation of vaccination of pregnant women and the types of circulating infl uenza viruses in relation to vaccine viruses, all aff ect vaccine effi cacy. If maternal vaccination is to succeed in Mali, infants born in September to October will need to be protected. Because northern hemisphere infl uenza vaccine becomes available in August or September, immunisation will need to be implemented almost immediately upon vaccine importation. The second peak (February to April) does not present this logistical issue. An alternative strategy for countries with an infl uenza epidemiology similar to Mali is to use vaccine with an extended shelflife throughout the year. 26 This approach would allow immunisation of Malian pregnant women in the months leading up to the September to October peak before the newer formulation is available.
Another factor infl uencing the overall eff ect and sustainability of maternal immunisation in countries such as Mali is access to health-care interventions. The 2014 Demographic and Health Survey reported that whereas 95·2% of pregnant women in Bamako and 91·8% in other urban areas had at least one prenatal visit during their most recent pregnancy, 27 this was true for only 69·3% of pregnant women in rural Mali. Barriers limiting access to health care in rural areas should be overcome to achieve high maternal immunisation coverage.
As other similarly low-resourced countries consider the implementation of a maternal infl uenza vaccination programme, the cost will also aff ect the feasibility. Since we did not observe an eff ect of trivalent inactivated infl uenza vaccine on birthweight, the cost-eff ectiveness of implementation of the vaccine in pregnancy to prevent infant infl uenza in Mali will hinge on prevention of severe illness and infant deaths. However, our study was not powered to measure the effi cacy of trivalent inactivated infl uenza vaccine in the prevention of severe laboratory-confi rmed infl uenza. Furthermore, because we visited households of study participants weekly, and intervened when illnesses were encountered by treating and referring participants earlier than they might have sought care in our absence, we probably interrupted progression of illness in many infants. Addressing this gap will require a diff erent trial design and a larger sample size.
Our study unequivocally demonstrates effi cacy of maternal immunisation against laboratory-confi rmed infl uenza among infants and mothers, and shows high acceptability and logistical feasibility, thereby paving the way for a larger trial to assess prevention of severe laboratory-confi rmed infl uenza leading to hospital admission in infants. Our fi ndings support a vision that, in the future, developing countries might use the maternal immunisation platform to deliver vaccines to prevent respiratory syncytial virus, 28 pertussis, 29 infl uenza, 13

Declaration of interests
We declare no competing interests.