A substantial proportion of the children in the Gambian pneumococcal vaccine trial had antibody concentrations considered to be protective against invasive pneumococcal diseases for most of the serotypes investigated at the age of 3–5 years, regardless of whether or not they had received PCV-9, although proportions were a little higher overall for children who had previously been vaccinated with PCV-9 and substantially higher for some serotypes. The relatively high concentrations of anti-pneumococcal antibodies observed in our study population are similar to those demonstrated previously in PCV vaccinated Gambian children 
. The significant differences in antibody concentrations prior to vaccination with PCV-7 for serotypes 6B, 14, and 23F between the groups indicate longevity of some vaccine induced antibodies. This may be a result of boosting by natural colonization and occult pneumococcal infection 
. However, it is of concern that antibody concentrations to serotype 1 polysaccharide were the lowest among those measured, as pneumococci belonging to this serotype are a frequent cause of IPD in The Gambia and in other developing countries. The high antibody concentrations in the control group for serotypes 4, 9V, 14 and 19F probably reflect past carriage with these organisms. High antibody concentrations to serotype 5, which was not commonly found in carriage studies of infants in the PVT 
nor in this follow-up study, may reflect past short-duration carriage which is less likely to be detected in cross-sectional studies, and/or past occult or symptomatic invasive infection. Previous studies of infants and toddlers in The Gambia have found carriage of serotypes 19F, 6B, 23F and 9V to be the most common among the vaccine-serotypes 
. Similar vaccine serotypes were found to be the most common in another carriage study that cuts across all age groups including adults prior to national routine PCV-7 vaccination in the Western Region of the Gambia 
. A high proportion of children with antibody concentrations of ≥0.2 µg/mL 5 years after primary vaccination with PCV-9 was described for serotypes 4, 6B, 9V, 18C, and 23F among HIV uninfected South African children 
; in our study proportions were lower for serotypes 1, 4, 18C and 23F than for other PCV-9-serotypes. These data support the existence of differential host and/or environmental factors influencing the responses to the various serotypes contained within pneumococcal conjugate vaccines.
A strong response to PCV-7 at the age of 3–5 years was seen in children in both groups. The proportion of children with antibody concentrations considered to be protective against invasive pneumococcal disease more than 3 years after primary immunization ranged from 55 to 99% in the PCV-9 group and 47 to 96% in controls, but the booster/delayed primary vaccination increased this proportion to >90% for all PCV-7 serotypes in both groups for up to 18 months, except for serotype 23F in controls which had fallen to 85% at 18 months post-PCV-7. These increases in the antibody concentration following booster/delayed primary PCV-7 contribute to consideration of whether or not to include a booster dose in the PCV vaccination schedule in countries where carriage of pneumococci is common. Such a review of vaccination schedules would require a study with more frequent assessment of the kinetics of antibody concentrations to fully assess a potential anamnestic response among previously vaccinated children. We have shown recently that in resource-poor settings, administration of a booster dose of pneumococcal polysaccharide vaccine (PPV-23) following primary immunization with one or two doses of PCV-7 diminished the differences in initial antibody responses and might lower the cost 
. Further study of optimal pneumococcal vaccination schedules would be helpful 
. The effect of a single dose of PCV-7 on antibody concentrations in the control group also supports the WHO recommendation that when PCV is first introduced into routine childhood immunization programmes a single catch-up dose of PCV-7 may be given to previously unvaccinated children aged 12–24 months and to children aged 2–5 years who are at high risk 
No significant differences were found in the proportions of children carrying pneumococci between the children who had received booster/delayed primary vaccination and the control group at any time point. Following booster or delayed primary immunization with PCV-7 there was a small decline in the prevalence of carriage with pneumococci of vaccine serotype as might have been expected, which was not seen for non-vaccine serotypes. The initial sample size calculation for this study was based on the assumption that the overall carriage rate of pneumococci of vaccine serotype in children in the control group would be 75%. Because this figure was only 17% there was minimal power to detect significant differences between groups. The increase in prevalence of carriage of pneumococci of non-vaccine type among children who had received PCV-9 in infancy previously reported 
persisted to the age of 3–4 years and up to 6 weeks after the booster vaccination, but this difference was minimal at and after 5 months post-PCV-7 vaccination.
Our study was not designed to assess correlation between antibody levels and carriage. Another report showed high levels of functional antibody after post-primary PPV-23 vaccination without impact on carriage, although there had appeared to be an effect of the number of doses of conjugate vaccine received on carriage at age 9 months 
. An earlier study of primary conjugate pneumococcal vaccination had found that higher IgG concentrations led to a decreasing probability of having a new acquisition of pneumococcal carriage of the corresponding serotype, and achieved statistical significance for serotypes 14 and 19F 
. Among adults, a pneumococcal anticapsular IgG concentration of 5 ug/mL has been shown to correlate with protection against carriage of serotype 14 
The prevalence of carriage before PCV-7 vaccination was lower than has been reported previously in other parts of The Gambia where overall carriage across a community including adults was 72%. It was 97% among children <1 year old and 93% among babies of ages <1 month 
. The lower than expected carriage of pneumococci before PCV-7 vaccination was surprising and may be in part related to temporal trends in pneumococcal carriage but also to the increasing age of the study population.
In conclusion, there were significantly higher antibody concentrations to 3 of the 9 serotypes in vaccinated children compared to controls approximately 3 years after primary vaccination with PCV-9, and antibody levels in PCV-9 recipients and controls were increased by PCV-7. Carriage of vaccine serotypes was low in both groups and we could not assess adequately the effect of PCV-7 on this endpoint.