Study Design and Oversight

The WHO Initiative for Vaccine Research coordinated the Measles Aerosol Vaccine Project, and the Centers for Disease Control and Prevention and the American Red Cross were partners in the project. An independent data and safety monitoring board had access to unblinded data to assess serious adverse events. A product development group (see the Supplementary Appendix, available with the full text of this article at NEJM.org) reviewed progress. Ethics committees of the WHO, Christian Medical College (in Vellore, India), the National Institute of Virology, and King Edward Memorial Hospital Research Centre (in Pune, India) approved the protocol. The trial was designed and conducted in accordance with Good Clinical Practice15 and Good Laboratory Practice16 guidelines.

The Serum Institute of India provided all vaccines free of charge, and Aerogen provided the delivery devices free of charge. Additional details about the trial design, conduct, and analysis are provided in the Supplementary Appendix and in the full protocol, including the statistical analysis plan, available at NEJM.org.

Study Participants and Clinical Setting

The trial was conducted in villages served by eight primary health centers in Pune. One study has shown that more than 90% of infants in Pune are breast-fed from birth, for a median of 4.7 months.17 Children between 9.0 and 11.9 months of age were eligible to participate in the study if they were due to receive primary measles vaccination. Children were excluded from participation if they were ineligible to receive measles vaccine according to WHO criteria.18

Randomization and Vaccination

From December 20, 2009, through April 5, 2010, we randomly assigned children, in a 1:1 ratio, to receive measles vaccine by means of aerosol inhalation or by means of a subcutaneous injection. A detailed description of the randomization process is provided in the Supplementary Appendix. After obtaining written informed consent from the parents or guardians of the children, the study nurses telephoned a centralized Web-based service and recorded the study assignments. At the Vadu site, which is part of a demographic surveillance system,19 two random subgroups of 100 children were selected to have blood drawn at either day 28 or day 364 for the monitoring of serologic responses.

We used a measles vaccine (Serum Institute of India), licensed by the WHO, that contained at least 1000 viral infective units of the live attenuated Edmonston–Zagreb strain of measles virus in each dose. The vaccine was delivered in 10-dose vials.

The study nurses reconstituted the 10-dose vials of measles vaccine and stored them until use at 2 to 8°C. They reconstituted the vaccine for aerosol delivery in 2-ml diluent and administered a single 0.2-ml dose, nebulized for 30 seconds through a single-use nonvented face mask, using a battery-operated Aeroneb vibrating mesh nebulizer (Aerogen). The nebulizer generated aerosol with a volume median diameter of 5.1 μm (geometric standard deviation, 2.1 μm) as determined by means of laser diffraction (Spraytec) (details are provided in the Supplementary Appendix).14 The study nurses reconstituted the vaccine for subcutaneous delivery in 5-ml diluent and administered a single 0.5-ml dose into the left upper arm. The rooms for delivery of aerosolized vaccine and subcutaneous vaccine were separate so that children receiving subcutaneous vaccine were not exposed unintentionally to aerosolized vaccine. Reconstituted vaccine was discarded after 6 hours.

End Points

Immunogenicity

The prespecified primary end point was seropositivity for serum antibodies against measles 91 days after vaccination. We defined seropositivity as 0.1 or more optical-density units on an enzyme-linked immunosorbent assay (ELISA) (Enzygnost Anti-Measles Virus/IgG, Siemens) or, in samples containing less than 0.1 optical-density units, a measles antibody concentration of 120 mIU per milliliter or more as measured with the use of a plaque-reduction neutralization test (PRNT). The testing algorithm was based on a study that showed a positive predictive value of 99.4% for ELISA (cutoff value, 0.1 optical-density units) as compared with PRNT (cutoff value, 120 mIU per milliliter).20

All samples at baseline and day 91 were tested by means of ELISA. Specimens with less than 0.1 optical-density units and all samples from the Vadu site were tested by means of PRNT. Paired prevaccination and post-vaccination samples were tested in the same run. All specimens were tested at the Virus Reference Department, Public Health England (formerly the United Kingdom Health Protection Agency), United Kingdom, in March 2012. The specimens had been tested in Pune, but the results of a random 10% sample of ELISA and PRNT analyses did not meet the prespecified quality-assurance criteria, so all samples were shipped to the United Kingdom.

Prespecified secondary end points were seroconversion and geometric mean concentrations of antibodies. Seroconversion was defined as a change from seronegative at day 0 to seropositive at day 91. Secondary outcomes were the difference in rates of seroconversion (among participants who were seronegative at baseline), the ratio of geometric mean concentrations, and geometric mean concentrations at days 0, 28, 91, and 364 in children in Vadu.

Safety

The prespecified primary safety outcome was the frequency of all solicited and unsolicited reports of adverse events up to 91 days after vaccination on day 0.21 Study nurses observed all children during and for 30 minutes after vaccination. Adverse events were then assessed according to the report of parents and guardians during home visits on days 3, 7, 10, 17, 21, 28, and 56 and by clinical examination on days 14 and 91 (Fig. S1 in the Supplementary Appendix). All children who were enrolled in Vadu were evaluated at day 364.

We collected information with the use of questionnaires that solicited information on 16 events (Table S1 in the Supplementary Appendix), by means of active surveillance both for events requiring treatment or hospitalization and for deaths, and through unsolicited reports of events from parents or guardians. We sent reports of all adverse events to the data and safety monitoring board, which had the authority to stop the trial if a single serious adverse event was judged to be attributable to the vaccine.

Follow-up and Blinding

We followed all participants until day 91 (from March through August 2010, depending on the date the patient underwent randomization). Case-based active surveillance with serologic confirmation of cases of fever and rash (Enzygnost Anti-Measles Virus IgM ELISA, Siemens) was in place in the entire trial area.22 Parents or guardians, children, and study staff were aware of the route of vaccine administration.

To reduce bias, all outcome assessments were blinded. Field workers used follow-up case-report forms that did not record the vaccination assignment, laboratory staff conducted analyses of coded specimens, and statisticians conducted data checks and preliminary analyses of blinded data.

Statistical Analysis

We aimed to show that seropositivity after receipt of aerosolized vaccine against measles was no more than 5 percentage points lower than seropositivity after subcutaneous vaccination. This estimate was based on the 4 percentage-point difference in a previous systematic review of studies involving children 10 to 35 months of age13; in addition, with a bigger margin, aerosolized vaccine would not provide levels of protection required for herd immunity.18

We assumed that exactly 90% of the children in each group would be seropositive at day 91, and we allowed for lower immunogenicity in children younger than 10 months of age.7 For noninferiority to be shown with the use of the confidence-interval approach, the lower limit of the two-sided 95% confidence interval for the difference (aerosol group minus subcutaneous group) in the proportion of seropositive children at day 91 had to be above −5 percentage points in the per-protocol population. If the upper confidence interval was below −5 percentage points, we would conclude inferiority, and if the lower interval was above zero, we would conclude superiority (Fig. S2 in the Supplementary Appendix). We estimated that with a sample of 800 children per group, the study would have 90% power to detect these differences. We planned to enroll 1000 children per group to allow for 20% who would not have results that could be evaluated at follow-up.

We calculated the difference between the proportion of children who were seropositive after receiving the aerosolized vaccine and the proportion who were seropositive after receiving the subcutaneous vaccine in the per-protocol population (which consisted of children who received the assigned vaccine, did not have any major protocol deviations, and had specimen results at day 91) and the full-analysis population (which consisted of all children who underwent randomization, excluding children for whom outcome data were missing). The full-analysis population was equivalent to a modified intention-to-treat population. We used the Wilson score method to estimate 95% confidence intervals.23 Multiple imputation was used to predict seropositivity when outcome data were missing, and we repeated the analysis with inclusion of all participants (see the Supplementary Appendix).24

For the secondary end points, we used logistic regression to investigate the association of prespecified factors with lack of seroconversion after receipt of aerosolized vaccine. For safety outcomes, we calculated the percentages of children (with 95% confidence intervals) who had any solicited or unsolicited reports of an adverse event or serious adverse event up to day 91.