from our population-based cohort study support that PCV7 has been effective in preventing IPD caused by vaccine serotypes in the context of a 2+1 PCV program in Denmark. PCV7-type IPD in <5 year-olds occurred mainly in unvaccinated children, with only seven cases in those who received at least one dose of PCV7 (less than 1 case per 100.000 person-years) and only three cases of true vaccine failures due to serotypes 14, 19F and 23F. Among these cases, one case of vaccine failure was observed in a severely immunosuppressed child following three PCV7 doses, and two cases were observed in immunocompetent children following two infant doses before they were eligible for their booster. In addition, none of the IPD cases caused by the additional PCV13 serotypes had been observed in children vaccinated with PCV13. Thus no PCV13-vaccine failures have been observed in the first 8-months after PCV13 was started to be used in Denmark.
Concomitantly with the decrease of PCV7 serotypes observed shortly after PCV7's introduction 
, the six additional serotypes included in PCV13 were found to be the predominant cause of pediatric IPD in our cohort of patients, similarly to what has been reported from other countries 
. However, following the introduction of PCV13 in the program, it is not clear from our data whether further benefit in preventing IPD in children under 5 years could potentially be achieved by including additional serotypes in a pediatric vaccine, since no additional serotypes have been shown to be clearly dominating the pneumococcal population causing replacement disease. This observation could be both related to differences in the invasive potential of pneumococci and characteristics of the population 
. Nevertheless, the surveillance period after PCV7/PCV13 introduction is relatively short (8 months), and serotype replacement has to be closely followed in the future in order to evaluate replacement disease in children.
During the surveillance period immediately following the introduction of PCV7, we found that nearly one-fourth of IPD cases (n
45/191) were caused by a serotype included in PCV7, but only 10 children were partially or completely vaccinated with PCV7 before IPD, and only 3 cases fulfilled criteria of vaccine failure (, ). Comorbid conditions were identified in one-third of vaccinated children with IPD due to PCV7 serotypes, regardless of the number of doses administered (). Similar results have been reported from the US, where children belonging to moderate and high-risk groups are recommended to receive PCV-vaccination 
. We have previously reported an approximated 8% prevalence of comorbid conditions in a nationwide cohort of patients with pediatric IPD in Denmark 
. However, the methods used for collecting data on comorbidity and estimating the prevalence of such conditions differ greatly between the two studies (by calculation of the Charlson index vs. telephonic interviews with physicians), which make direct comparison of the findings difficult. In any case, our findings underline the importance of continued vigilance for pneumococcal disease in febrile children with comorbidity or other risk factors that make them more susceptible to the disease even after immunization with PCV.
The first case identified as a vaccine failure was due to a serotype 14 infection, a serotype which has been otherwise considered to be quite immunogenic 
. A case of serotype 14-vaccine failure has previously been described in a patient enrolled in the efficacy trial carried out among Navajo and White Mountain Apache children in the US 
. In contrast to this case, our patient was severely immunosuppressed before IPD onset. In the case from the US, the child was not known to have underlying medical conditions and had a high concentration of antibodies against serotype 14, which were also found to be functional in the OPA assay. The authors suggest that serotype-specific correlates of protection based on the ELISA cut-off values from population-based vaccine efficacy studies are not necessarily adequate to predict the level of protection at the individual level. Due to the critical clinical condition of the patient in our cohort, it was considered inappropriate to collect samples for immunological testing in this patient and they would probably be of limited value.
Serotype 19F has been found to be the most frequent serotype causing pneumococcal meningitis in PCV7-vaccinated children in France 
. In spite of its good immunogenicity, serotype 19F has been described to induce antibodies of lower avidity that decline faster compared to other antibodies induced by the vaccine 
. A case of vaccine failure due to serotype 19F has previously been described in the literature, in a child who presented with pneumococcal meningitis in spite of three PCV7 doses and who was considered immunocompetent 
. In our case, the clinical course was not severe; the child was discharged shortly after admission to the hospital with oral antibiotics. The clinicians decided not to investigate the antibody profile after IPD and had no clinical suspicion of immunosuppression in this patient.
Although the infant we described with 23F vaccine failure had received only 2 doses of PCV7, the antibody response to all the vaccine serotypes except serotype 23F was satisfactory, both measured by ELISA and OPA assay (). The child was considered to be immunocompetent. A lower vaccine response to this serotype has previously been reported, with the protective threshold for antibodies to this serotype (alike serotype 6B) being reached only after the booster dose 
Recommendations regarding clinical management and revaccination of children presenting vaccine failures must be tailored to the individual child in different clinical settings. Probably, an additional vaccine-dose followed by antibody responses and OPA measurements are suitable recommendations, since it is well documented that in some cases of vaccine failures the patient remains refractory to responses to the invading serotype despite further doses after the infection 
Concerning IPD caused by PCV7 serotypes in non-vaccinated patients, we found that nearly 40% of children were eligible for vaccination but presented with IPD before the first dose, mainly children who were targeted by the catch-up program. It has also been described that the coverage of PCV7 is slightly lower than the coverage of DTaP-IPV/Hib, and particularly for the catch-up program. The reasons for this difference are currently unknown 
. Still, PCV7's uptake achieved in Denmark is one of the highest in Europe, where large differences in PCV7 vaccination schemes and use of PCV7 between regions have been reported 
Among the strengths of our study, we can mention that we were able to report high-quality data collected through an enhanced surveillance system based on collaboration between NSR, the Department of Epidemiology, the departments of clinical microbiology and clinicians across the country (). The register of pediatric IPD due to PCV7 serotypes or to one of the six serotypes included in PCV13 includes complete clinical, epidemiological and laboratory information on patients, which will allow us to closely monitor and improve our understanding on cases of vaccine failure. Among the needs and opportunities for our surveillance system in the future we envisage the collection of information on vaccination history from IPD-cases due to other serotypes than those included in the vaccines. This would allow us to investigate whether certain groups of children are at higher risk of getting IPD and directly assess the effectiveness of the vaccines using the indirect cohort method 
The results of our study may be limited by several factors. We present the results from a relatively short surveillance period (39 months) which followed the introduction of PCV7 in the immunization program. This may be of importance in terms of evaluation of replacement disease among vaccinated children, considering that pneumococcal populations are highly dynamic over time and susceptible to external factors such as conjugate vaccination pressure 
. Also, children in the study population have recently been vaccinated with PCV and a large part of the population is still incompletely vaccinated, limiting the extrapolation of these results in the long term.
A selection bias may also be present in our study, since we used a highly specific case definition including only culture confirmed IPD cases in a setting were blood- and CSF-cultures are almost exclusively obtained in hospital settings. Furthermore, we were not able to investigate the immunological status after IPD in all children with vaccine failure, which could have contributed to the understanding the underlying mechanisms.
In summary, IPD in children younger than 5 years of age was a rare event and occurred mainly in unvaccinated children in our cohort of patients. Only three cases of PCV7-failure were confirmed during a 31-month surveillance period. Seven additional IPD cases due to PCV7- serotypes were observed in incompletely vaccinated children. So far, no PCV13-vaccine failure cases have been observed, however the surveillance period was short. Children with confirmed or suspected immunosuppression should be closely monitored and assessed on an individual basis.