Invasive Haemophilus influenzae Disease, Europe, 1996–2006

Incidence and case-fatality ratios are higher for non–type b than for type b infection.

The introduction of the Hib conjugate vaccine into national childhood immunization programs in the 1990s has resulted in a marked and sustained reduction in the incidence of invasive Hib disease in many countries (2). However, concern exists about the long-term effectiveness of the Hib immunization programs (12) and possible disease replacement by other H. infl uenzae strains (13). Because Hib conjugate vaccine reduces pharyngeal carriage (14), other H. infl uenzae strains theoretically could take its place and cause invasive disease (13,15). Elimination of Hib carriage also may reduce natural boosting of immunity, thereby resulting in lower Hib antibody and increased susceptibility to invasive Hib disease in the long term (16).
In 1996, a collaborative surveillance network was established in Europe to describe the impact of routine Hib vaccination on the epidemiology of invasive H. infl uenzae disease. By 2006, a total of 28 countries participated in surveillance, and 14 countries, comprising an annual denominator population of 150 million persons, routinely serotyped all invasive clinical H. infl uenzae isolates. We describe the epidemiology of invasive H. infl uenzae disease in countries with established national Hib immunization programs, devoting particular attention to invasive non-type b H. infl uenzae disease.

Methods
A 3-year European Union-funded study (the BIOMED II Hib surveillance project) was initiated in 1996 to study the epidemiology of invasive H. infl uenzae after the introduction of the Hib conjugate vaccine into national infant immunization programs. In 1998, this program was renamed European Union Invasive Bacterial Infection Surveillance (EU-IBIS) and expanded to include more countries; by Invasive Haemophilus infl uenzae Disease, Europe, 1996-2006 2006, a total of 28 countries routinely reported cases to EU-IBIS (www.euibis.org). Participating countries reported cases of invasive H. infl uenzae disease to a central database, with basic demographic details, clinical syndrome, outcome, specimen site, and serotyping data on isolates. The UK Health Protection Agency Haemophilus Reference Unit (HRU) coordinated development of standardized laboratory protocols for growing, serotyping, and PCR genotyping H. infl uenzae and exchange of clinical isolates to ensure consistency of results. HRU also provided genotypic confi rmation of serotypes for countries without established reference facilities and regularly distributed quality assurance strains to participating laboratories to ensure comparability of results. Annual reports were made available to participants at meetings and on the EU-IBIS website. Invasive H. infl uenzae disease was defi ned as isolation of the organism from a normally sterile site. A case of meningitis was defi ned as H. infl uenzae cultured from cerebrospinal fl uid or clinical and/or radiologic features of meningitis with blood culture positive for H. infl uenzae. Other clinical presentations, including epiglottitis, pneumonia, cellulitis, and osteomyelitis, were defi ned as isolation of H. infl uenzae from a normally sterile site (usually blood cultures but occasionally from another sterile site, e.g., joint fl uid in septic arthritis or pleural fl uid in empyema) in a person with clinical signs and symptoms consistent with that presentation. Bacteremia was defi ned as growth of H. infl uenzae from blood cultures only, with no distinctive clinical syndrome identifi ed.
This study included only countries that routinely vaccinated children against Hib before 2000 and serotyped at least 50% of all clinical isolates. Detailed surveillance methods for all participating countries are available at www.euibis. org. Germany and Israel reported data for children only and were included in some of the analyses. Within the United Kingdom, data from England and Wales were collected separately from Scotland because the surveillance program in England and Wales is separate from that in Scotland. In Greece, during 1996-2002, surveillance was limited to a single prefecture, Attiki, and provided data only for persons <15 years of age; after 2002, national data were available for Greece. Italy initially relied solely on laboratory reporting with voluntary notifi cation of confi rmed cases of meningitis, but a more active laboratory-based surveillance system was established in 8 Italian regions in 1997-1998 (Campania, Liguria, Lombardia, Piemonte, Puglia, Toscana, Trento, and Veneto), 7 regions in 1999-2002 (Lombardia was no longer included), and nationally thereafter. All data were collected as part of enhanced national surveillance and rendered anonymous at the source.
Data were entered by using Microsoft Excel (Microsoft, Redmond, WA, USA), and statistical analysis was performed by using Stata 8.0 (www.stata.com). Total and age-grouped population estimates used as denominators for incidence calculations were obtained either from the national statistics website of the relevant country or from EU-IBIS participants (17). The denominator for disease incidence in infants <1 month, 1-5 months, and 6-11 months of age was estimated by dividing the number of infants (<1 year of age) by 12 and multiplying by the number of months in each age group, respectively. For non-type b encapsulated H. infl uenzae infections, we combined data on children 5-14 years of age with data on adults 15-64 years of age because serotype distribution, clinical presentation, and outcomes were similar for these 2 age groups. To estimate any increase in incidence of non-type b H. infl uenzae disease during the study period, an overdispersed Poisson regression model was fi tted for the number of non-type b cases with covariates for year and country. To allow for changes in population and proportion of total H. infl uenzae isolates serotyped by country and year, these variables were included in the model as offsets. To control for differences in the collection of death data, we estimated case-fatality ratios (CFRs) by using the number of reported deaths as the numerator and all cases, including those for which outcome was not reported, as the denominator. We calculated age-adjusted odds ratios (ORs) for noncapsulated H. infl uenzae (ncHi) and Hib using logistic regression; we inclulded age and serotype (e.g., ncHi vs. Hib, H. infl uenzae type e [Hie] vs. H. infl uenzae type f [Hif]) as independent variables. Ages are given as medians and interquartile ranges (IQRs) and compared by using the Mann-Whitney U test. Proportions were compared by using the χ 2 test or Fisher exact test; continuous variables were compared by using Student t tests.

Results
During 1996-2006, a total of 14 countries reported 10,081 invasive H. infl uenzae cases, of which 2,836 (28.1%) were Hib, 690 (7.8%) were other capsulated H. infl uenzae, and 4,466 (44.3%) were ncHi. For 125 (1.2%) cases, the isolates were identifi ed as non-type b, but complete serotyping was not performed. Capsular serotype was not available for 1,964 (19.5%) isolates, mainly because the isolate was not available for typing at the reference laboratory (e.g., where the isolate could not be recultured). The crude overall annual incidence rates for invasive Hib, ncHi, and non-type b encapsulated H. infl uenzae infections were 0.15, 0.28, and 0.036 cases per 100,000 population (Table 1). After adjusting for the proportion of isolates not serotyped in each country per year and population changes over time, we found a small but statistically signifi cant increase in the incidence of non-type b H. infl uenzae disease ( (Figure). More than half of Hib cases, compared with fewer than one quarter of ncHi cases, occurred in children <5 years of age (χ 2 = 438; p<0.0001). By contrast, only 13.7% of Hib cases occurred in persons >65 years of age, compared with 42.7% of ncHi (χ 2 = 483; p<0.0001). The most common clinical diagnosis was bacteremia for both Hib and ncHi, but the median age of persons with Hib bacteremia was much lower than for that for those with ncHi bacteremia (Table 2). Hib meningitis occurred mainly in infants, whereas ncHi meningitis occurred in all age groups. Overall, ncHi was responsible for one third of all meningitis cases but accounted for 18.4% of meningitis cases in children <5 years of age and for 62.9% of persons >65 years of age. Hib was responsible for 84.1% of epiglottitis cases; only 6.6% were caused by ncHi. Hib pneumonia occurred mainly in adults and elderly persons; ncHi pneumonia occurred more often in children <5 years of age and in elderly persons.
Invasive infections caused by non-type b encapsulated H. infl uenzae were rare. Cases did not cluster by country or time, and individual serotypes or incidence, either overall or in any single country, did not increase during the 11-year study period. Of the 690 cases, Hif was the most prevalent subtype (500 [72.5%] patients) followed by Hie (143 [20.7%]) ( Table 4). The overall CFR was 9.1% (63/690 patients). The CFR increased with age and was highest for bacteremia (

Discussion
The marked reduction in invasive Hib disease after the introduction of the Hib conjugate vaccine had prompted concerns that other H. infl uenzae serotypes, ncHi, or other respiratory pathogens might fi ll the ecologic niche. However, little evidence exists for a substantial or sustained increase in invasive non-type b H. infl uenzae infections (18). The rise in Hib incidence during 2000-2002 resulted mainly from an increase in the United Kingdom and the Netherlands; however, rates remained well below those in the prevaccine era (19,20).
Although prospective enhanced national surveillance may be incomplete (21), comparisons over time and across serotypes are largely valid, assuming serotyping is accurate and complete. In addition, although lower ascertainment might lead to lower estimation of the true incidence of invasive H. infl uenzae disease, it is less likely to affect the clinical presentation, age distribution, outcome, or proportion of cases due to the different serotypes. All participating countries had reference laboratories that routinely serotyped all invasive H. infl uenzae strains and participated in an external quality assurance scheme. As a result, 80.5% of 10,081 H. infl uenzae isolates identifi ed were serotyped. The robustness of the surveillance is demonstrated by the incidence of inva-sive non-type b disease, which remained relatively constant over the 11-year study period despite increasing numbers of participating countries and provides further confi dence that replacement disease is not occurring (18,22,23).
In countries with established Hib immunization programs, the incidence of ncHi is now higher than that of Hib. Unlike Hib, however, invasive ncHi infections occur mainly in neonates and elderly persons. Neonatal ncHi infections are well described but account for <5% of all neonatal invasive bacterial infections (3,24,25). The infection develops rapidly (usually within 48 hours after birth) and follows a fulminant course with a high CFR, particularly in preterm infants (3). Invasive ncHi disease in neonates also is associated with septicemia in the mother, increased complications during labor, and preterm delivery (24,26,27). In our study, ncHi was more common among women in  the 25-44-year age group, suggesting that childbearingaged women may be at increased risk for invasive ncHi infections. This fi nding may refl ect increased exposure, for example, because of contact with children or increased susceptibility, such as in pregnancy. In older children and adults who develop invasive ncHi infections, studies have reported that more than half the case-patients had serious predisposing medical conditions, such as chronic respiratory disease and impaired immunity (3)(4)(5)(6)(7). Unfortunately, because clinical information collected for individual cases in our study was limited, we could not further elucidate possible risk factors for invasive infections caused by the different serotypes.
Infection from non-type b encapsulated H. infl uenzae is extremely rare and mostly caused by Hif and Hie. Other population-based studies also have reported a predomi-nance of Hif and, to a lesser extent, Hie among non-type b encapsulated H. infl uenzae in adults and children (3,6,28). The clinical presentations of both Hif and Hie disease are almost identical and similar to that of ncHi infections in that almost half the cases occurred among persons >65 years of age who usually had bacteremia and pneumonia (6,7). Although 44% of Hif infections occurred among persons >65 years of age, compared with 21% among children <5 years, the incidence was almost the same in the 2 age groups. Hif and Hie have considerably restricted genetic diversity, and most infections are caused by a few strains that may be intrinsically more pathogenic than noninvasive strains (29). Other studies have reported that, as with ncHi, 60%-80% of persons with invasive Hif disease and Hie had underlying conditions that predisposed them to opportunistic infections (3,6,(9)(10)(11)25,(30)(31)(32).     In contrast to Hif and Hie, invasive Hia infections were similar to Hib infection in that they occurred mainly in young children who often had meningitis (2). In our study, the incidence of Hia in children <5 years of age (0.12/million) was much lower than that reported in Navajo and White Mountain Apache children <5 years of age (20 cases/100,000 population) (33), Alaska Native children <2 years of age (21/100,000) (34), and northern Canadian aboriginal children <2 years of age (102/100,000) (34). The same populations are also highly susceptible to invasive Hib disease (35,36). Hia and Hib have the most closely related capsules (37) and a similar degree of genetic diversity (29).
Infections caused by Hic and Hid are rare and have low CFRs, suggesting that they may not be particularly virulent.
There is a paucity of information on infections caused by these serotypes, even in the form of individual case reports. Our data suggest that these invasive Hic and Hid infections are more common in adults. A recent US study reported that, of 770 cases of invasive H. infl uenzae disease during 1996-2004 in Illinois, 3 (43%) of 7 Hic, and 4 (67%) of 6 Hid cases occurred in persons 18-64 years of age (28).
That CFR from invasive Hib disease remains low and similar to that reported in other industrialized countries is reassuring, even though it has not changed substantially from the prevaccine era (2,38). In contrast, CFRs for ncHi and non-type b encapsulated H. infl uenzae were significantly higher than for Hib. The CFRs in our study should be considered a minimum because we cannot be sure that all deaths would be reported to the surveillance systems, particularly if death occurred a considerable time after infection. Other studies with more active follow-up in adults with ncHi infections have reported CFRs of 13%-20% (28) and up to 29% within 1 month after infection (7). Although the high CFRs associated with early-onset neonatal ncHi is well described (3), our fi nding of such high CFRs in infants was unexpected. Whether these infants had any underlying medical conditions that predisposed them to death or the organisms causing infection in this age group are more virulent is not known. The CFR for invasive non-type b encapsulated H. infl uenzae infections was also higher than for Hib and comparable to the 15%-30% reported in other studies (7,28). The higher CFR for invasive ncHi infections among elderly persons and persons with other clinical diagnoses (Table 3) most likely results from a higher prevalence of underlying medical conditions predisposing them to opportunistic infections. In the latter group, for example, ncHi was often isolated from uncommon sterile sites, such as peritoneal and pericardial fl uid, renal and spleen biopsy specimens, and brain abscesses, suggesting that such persons may have serious underlying medical conditions at the time of infection. Underlying medical conditions also may explain why CFRs may be higher for persons with invasive Hie infections; the small number of Hie cases compared with Hib, ncHi, or Hif suggests that this serotype is not particularly virulent. Other studies have reported higher CFRs for Hie than for Hif, particularly for elderly persons (28). Thus, despite the reduction in Hib disease, continued surveillance is needed for all H. infl uenzae infections across all age groups to assess the long-term effectiveness of Hib vaccination, rapidly detect unexpected population effects and potential changes in circulating strains, and monitor changes in the epidemiology of invasive H. infl uenzae disease. Further studies are needed to defi ne more clearly host and pathogen risk factors for invasive H. infl uenzae infection and factors associated with death.