Conjugate vaccine introduction in the African meningitis belt: meeting surveillance objectives
Corresponding Author Judith E. Mueller, Department of Epidemiology and Biostatistics, EHESP School of Public Health, Hôtel-Dieu, 1 place du Parvis Notre-Dame, 75181 Paris, France. E-mail: email@example.com
The epidemiology of bacterial meningitis will change with introduction of meningococcal and pneumococcal conjugate vaccines in the African meningitis belt. The principal objectives of surveillance are to evaluate the impact of vaccination, to detect and investigate epidemics and provide material for research. The capacity of existing surveillance activities in the meningitis belt to meet these objectives varies due to infrastructural and financial constraints. Impact assessment of conjugate vaccine against meningococcal serogroup A will be limited to comparing incidence trends from a few surveillance sites with data obtained before vaccine introduction and to comparing trends in the incidence of suspected cases or localised epidemics in most other settings. The timeliness of detection of epidemics and identification of epidemic meningococcal serogroups could be improved in most countries by analysing suspected case data in health centre level resolution and by investigating outbreaks with mobile teams. For research and impact assessment of pneumococcal conjugate vaccines, several surveillance sites covering at least 0.5 million inhabitants should be maintained which undertake exhaustive case finding and systematic laboratory confirmation of meningitis and pneumonia. Molecular diagnostics will facilitate surveillance in remote areas, but the available techniques should be evaluated for diagnostic performance in the field and long-term sustainability.
L’épidémiologie de la méningite bactérienne va changer avec l'introduction des vaccins conjugués contre le méningocoque et le pneumocoque dans la ceinture africaine de la méningite. Les principaux objectifs de surveillance sont d’évaluer l'impact de la vaccination, de détecter et enquêter sur les épidémies et de fournir du matériel pour la recherche. La capacité des activités de surveillance existantes dans la ceinture de la méningite pour atteindre ces objectifs varie en raison de contraintes infrastructurelles et financières. L’évaluation de l'impact du vaccin conjugué contre le méningocoque du sérogroupe A sera limitée à la comparaison des tendances de l'incidence dans quelques sites de surveillance avec les données obtenues avant l'introduction du vaccin et de comparer les tendances de l'incidence des cas suspectés ou des épidémies localisées dans la plupart des autres endroits. La rapidité de la détection des épidémies et de l'identification des sérogroupes méningococciques dans les épidémies pourrait être améliorée dans la plupart des pays en analysant les données sur les cas présumés au niveau du centre de santé et en recherchant les épidémies avec des équipes mobiles. Pour l’évaluation de la recherche et de l'impact des vaccins antipneumococciques conjugués, plusieurs sites de surveillance couvrant au moins 0,5 millions d'habitants, devraient être maintenus pour entreprendre la recherche exhaustive des cas et la confirmation systématique en laboratoire de la méningite et la pneumonie. Les diagnostics moléculaires faciliteront la surveillance dans les zones reculées, mais les techniques disponibles doivent être évaluées pour la performance diagnostique sur le terrain et à long terme.
La epidemiología de la meningitis bacteriana cambiará con la introducción de las vacunas conjugadas para meningococo y neumococo en el cinturón Africano de meningitis. Los objetivos principales de la vigilancia son evaluar el impacto de la vacunación, detectar e investigar epidemias, y proveer material para investigación. La capacidad de las actuales actividades de vigilancia en el cinturón de meningitis para alcanzar estos objetivos, varía debido a limitaciones financieras y de infraestructura. La evaluación del impacto de la vacuna conjugada contra meningococo del serogrupo A estará limitada a comparar las tendencias en la incidencia de unos pocos emplazamientos bajo vigilancia, con datos obtenidos antes de la introducción de la vacuna, y a comparar las tendencias en la incidencia de los casos sospechosos o las epidemias localizadas en la mayoría de los otros emplazamientos. La detección a tiempo de las epidemias y la identificación de los serogrupos epidémicos de meningococo se podría mejorar en la mayoría de los países, analizando datos de casos sospechosos a nivel de centros sanitarios e investigando brotes epidémicos con equipos móviles. Para investigar y evaluar el impacto de las vacunas neumocócicas conjugadas, deberían mantenerse varios emplazamientos bajo vigilancia cubriendo al menos 0.5 millones de habitantes, llevando a cabo una búsqueda exhaustiva de casos y una confirmación sistemática de laboratorio para meningitis y neumonía. El diagnóstico molecular facilita la vigilancia en áreas remotas, pero las técnicas disponibles deberían evaluarse en lo que respecta al desempeño diagnóstico en el campo y la sostenibilidad a largo plazo.
Epidemic meningitis has been observed in the African meningitis belt for over a century (Greenwood 1999). Initially, surveillance was limited to counting suspected cases and deaths (Lapeyssonnie 1963), but progress in microbiological techniques led to aetiology-specific surveillance (Broome et al. 1983) and observation of strain epidemiology (Nicolas et al. 2005). Thresholds were set for the incidence of suspected cases to guide epidemic vaccine response (WHO 2000). Introduction of a conjugate vaccine against Haemophilus influenza type b (Hib) is now being followed by MenAfriVac®, a conjugate vaccine against Neisseria meningitidis serogroup A meningococci (NmA) and pneumococcal conjugate vaccines (PCV) against 10 or 13 serotypes. This presents new challenges for surveillance, as the long-term justification for vaccine requires evidence of impact, and optimisation of vaccination strategies requires specific incidence data; furthermore, the vaccines do not cover all serogroups and serotypes. Specific surveillance strategies are needed in the meningitis belt, where meningococcal meningitis occurs due to various serogroups in localised and regional epidemics, and the incidence of meningococcal and pneumococcal meningitis show 10- to 100-fold seasonal incidence variations (hyperendemicity) (Mueller & Gessner 2010). Persisting infrastructural barriers in this sparsely populated, vast geographical region where most people live in rural areas represent a particular challenge. This paper discusses how the epidemiology of meningitis and the objectives and funding of surveillance are changing with the introduction of conjugate vaccines and how surveillance strategies could be adapted from an epidemiological point of view.
What will change?
Epidemiology of meningitis
Mass vaccination campaigns with MenAfriVac® targeting the population aged 1–29 years will cover the entire meningitis belt and bordering countries between 2010 and 2016 (LaForce & Okwo-Bele 2011). This catch-up strategy will ultimately be supplemented by routine infant vaccination to avoid the accumulation of susceptible children. MenAfriVac® is expected to provide indirect protection by reducing NmA colonisation and transmission, and ultimately lead to elimination of NmA epidemics (LaForce & Okwo-Bele 2011) and a substantial overall reduction in bacterial meningitis incidence. Threats to these outcomes include limited vaccine coverage, reduced vaccine effectiveness in the specific situation of the meningitis belt and the epidemic emergence of other serogroups. Furthermore, NmA transmission could persist among older age groups who carry NmA at substantial prevalence, at least during epidemics (Hassan-King et al. 1979; Mueller et al. 2011), but who were not targeted by vaccination; this could maintain hyperendemic incidence or even outbreaks in these age groups. In addition, the duration of antibody persistence is still unknown and waning mucosal immunity could allow the re-establishment of NmA circulation a few years after vaccine introduction, followed by accumulation of population groups susceptible to disease. In consequence, NmA meningitis would progressively resume seasonal hyperendemicity with eventual occurrence of epidemics. Finally, there is a hypothetical risk of serogroup replacement, in which case hyperendemic and epidemic meningitis incidence due to other serogroups would increase after vaccine introduction. While these are worst-case scenarios, surveillance systems should be capable of rapidly detecting such events, for epidemic response and for adaptation of the vaccination strategy.
Introduction of PCV in infant immunisation programmes in the meningitis belt, already achieved in Mali, will be accelerated by financial support from the Global Alliance for Vaccines and Immunisation (GAVI). Routine immunisation with PCV in addition to MenAfriVac® is expected to reduce the hyperendemic incidence of bacterial meningitis at least in the target age groups to levels comparable to those on other continents.
Objectives of surveillance
The prime objective of surveillance over the next few years will be to evaluate the impact of MenAfriVac® and PCV in several if not all countries in the meningitis belt. Evidence that the vaccine has an impact is essential to attract continued economic investment in preventive vaccination programmes (WHO 2007). The impact, efficacy and effectiveness of MenAfriVac® are yet to be documented, as the vaccine was licensed on the basis of its immunogenicity only (Sow et al. 2011). The situation is similar for PCV, where the results of efficacy trials of conjugate vaccines against pneumococcal serotype 1, the principal pathogenic serotype in this region (Gessner et al. 2010), have yielded ambiguous results (Cutts et al. 2005; Klugman et al. 2003).
Epidemic detection and investigation will remain objectives of surveillance, as they guide the vaccine response and the support to patient management. Vaccine response is effective, however, only if implemented early during an epidemic curve (Leake et al. 2002) and against the serogroup responsible for the epidemic. These factors remain challenges, particularly as there is no vaccine against serogroup X.
Surveillance also provides the material for research on the meningitis belt phenomenon, associated diseases and future vaccine development. It serves to identify representative patient samples for detailed clinical and biological investigations, provides incidence estimates for interpretation of the results of carriage and immunological studies, and the outcome variable in risk factor or modelling studies. This objective of surveillance will retain its supplementary character in the future.
As declared by published reports, funding for meningitis surveillance during the past decade came from various sources, including meningococcal vaccine manufacturers, foreign and local governments, international organisations, private foundations and vaccine initiatives like the Meningitis Vaccine Project (Leimkugel et al. 2005; Siéet al. 2008; Massenet et al. 2009; CDC 2009; Collard et al. 2011; Delrieu et al. 2011; WHO-Afro 2012). Several surveillance sites are threatened or have ceased activity due to lack of funding and little funding may be available for surveillance in the future; particular reasons may include the fact that the public health problem of epidemic meningitis is assumed to be solved (Roberts 2010) and that surveillance, unlike other interventions, documents the burden of disease but does not reduce it and thus is less attractive than intervention programmes. Nevertheless, strategic investment in surveillance for an extended time is required for valid, representative assessment of the impact of current and future vaccines in the meningitis belt.
Requirements for surveillance
Requirements for accurate surveillance of vaccine-preventable diseases have been well defined (Chen & Orenstein 1996), but may vary in importance according to objectives and context.
Assessing impact from routine surveillance data is challenging as the switch from epidemic alert to benefit analyses places ‘new demands on the quality of the data and on the stringency of epidemiological analysis’ (Giesecke 2002). Vaccine impact is usually assessed by quantifying the reduction in disease burden by analysing trends before and after vaccine introduction, or by stepped wedge design for comparison with a control group. For bacterial conjugate vaccines in the meningitis belt, the population surveyed before and after introduction should comprise at least 0.5 million inhabitants. This recommendation is based on a sample size calculation for estimating an incidence rate ratio between the periods before and after vaccine introduction, with 95% confidence intervals that do not include 1. For example, 100 000 people aged 15–29 years must be surveyed to show significance for a 50% decrease in the annual incidence of any meningococcal meningitis from 30 (Campagne et al. 1999) to 15 cases per 100 000 per year. This age group represents about 17% of the population of Burkina Faso (2003 census data) and a total population size of 588 000 is thus required.
Bias should be avoided by using the same protocol for the periods before and after vaccine introduction and by exhaustive case finding in the community; surveillance in all health centres treating bacterial meningitis in the population may suffice, if attendance patterns are stable over time. The laboratory tests used should have high diagnostic performance or at least a well-evaluated and stable performance profile. The basic units for comparison of trends are annual incidence rates calculated for full years, as monthly rates mainly reflect pronounced seasonal variation. The observation period both before and after vaccine introduction should be 10 years or more in order to cover full cycles of NmA epidemic waves (Greenwood 1999; Mueller & Gessner 2010).
Alternative measures of vaccine impact include a decrease in the incidence of localised epidemics at the health centre level (Tall et al. 2012a). While the absence of NmA meningitis incidence in vaccinated populations and high incidences in neighbouring unvaccinated populations suggests an impact, such comparisons are frequently biased by natural variations in the circulation of strains and by differing population characteristics and surveillance methods.
Evidence for the effectiveness of a vaccine from case-control studies confirms its preventive potential. If valid estimates of vaccine coverage and disease incidence are available, such estimates of vaccine effectiveness can be extrapolated to determine the approximate vaccine impact in a population. This approach has the advantage of not requiring historical data, but cannot be used if vaccine introduction led to complete elimination of NmA meningitis, for example due to herd immunity or during periods of natural absence of NmA cases (Soriano-Gabarróet al. 2007), which may have been the case in most countries of the meningitis belt during 2011–2012 (WHO-Afro 2012).
For PCV impact assessment, serotype-specific meningitis incidence estimates are required. As the seasonal variation (Leimkugel et al. 2005) of pneumococcal meningitis incidence is similar to that of meningococcal meningitis but the pluri-annual variation is minor, trends should be compared from annual incidences, while a shorter observation period of a few years before and after introduction is sufficient. Surveillance should cover pneumonia, which is the disease associated with most pneumococcal infections. Given the high risk for pneumococcal meningitis throughout adulthood (Gessner et al. 2010), surveillance for PCV impact assessment should include adolescents and adults.
Detection of epidemics
Surveillance for detecting meningococcal epidemics ideally involves almost complete geographical coverage, is exhaustive and stable over the comparison periods and ensures rapid data transmission and analysis. Systematic laboratory confirmation and the linking of biological and clinical data are secondary in routine. By contrast, once an epidemic signal is detected, rapid laboratory confirmation including serogrouping is essential for an appropriate vaccine response.
For research, the important characteristics of surveillance are the representativeness of the surveyed population for the larger region, knowledge of the source population size to allow calculation of incidence rates, completeness of case finding and reporting, linking of clinical and biological information, use of high-performance diagnostic tools and the inclusion of all relevant clinical forms of the infection. Geographical coverage and sustainability over extended periods are of lesser importance.
Although clinical decision-making is not within the scope of this article, it is affected by surveillance. For example, surveillance data will guide adaptation of treatment protocols in settings with no reliable clinical laboratory diagnosis, and adaptation of the positive predictive value of diagnostic tests according to changing probability of disease. In remote districts, surveillance test results are often the only diagnostic tool available, and clinicians’ decision may be made on the basis of, e.g. the appearance of cerebrospinal fluid (CSF) after lumbar puncture, which has poor sensitivity for bacterial meningitis (Rose et al. 2010), or a rapid test for detecting meningococcal serogroups, which gives a negative result for pneumococcal or serogroup X meningitis (Rose et al. 2010). For improved clinical management in remote settings, reliable bedside tests are needed that allow a distinction between meningococci, pneumococci and other pathogens, and that ideally detect antibiotic resistance.
Current surveillance systems
Reporting of suspected cases and enhanced meningitis surveillance
Most countries in the meningitis belt have a system for weekly reporting of suspected meningitis cases with high geographical coverage and completeness of reporting (WHO-Afro 2011). Data aggregated for each district are available for the past 20 years (Thomson et al. 2006). This system does not include microbiological confirmation and is therefore not costly, but it requires investment by countries in training and data management, and coordination by WHO. It allows evaluation at national or district level of whether the introduction of conjugate vaccines has decreased the average annual or seasonal variation in the incidence of suspected meningitis. Its main limitation is that trends such as the low incidence in 2011 in most countries of the meningitis belt (WHO-Afro 2011) cannot be attributed to specific pathogens. Health workers’ alertness to bacterial meningitis might vary by season and according to vaccine introduction, leading to biased temporal trends. Reporting is quite exhaustive in most settings, although underreporting of infant and foudroyant cases is likely. It is a good tool for epidemic detection, as the thresholds are sensitive and relatively specific (Leake et al. 2002). Data analysis in health centre resolution (i.e. with the health centre, not the district, as the basic observation unit) is currently recommended only during epidemics in order to localise the herd within a district (WHO-Afro 2009). Data from these surveillance systems are limited for research purposes by their non-specificity for individual pathogens and, for pneumococcal research, the absence of standardised case definitions for reporting lower respiratory tract infections.
To complement these systems, the Enhanced Meningitis Surveillance protocol is used for microbiological investigation of suspected cases. In its current version, as a part of Integrated Disease Surveillance and Response (WHO-Afro 2009), it includes latex agglutination testing in district laboratories for rapid serogrouping, CSF transport in trans-isolate media for culture analysis at regional (without serogrouping) and reference (with serogrouping) laboratories, transport in dry tubes for polymerase chain reaction (PCR) at national reference laboratories, collection of information on vaccination status and socio-demographic data and unique identifiers for linking clinical and laboratory data. Although the protocol fulfils most of the criteria for ideal surveillance, the actual completeness and timeliness are not satisfactory in most countries (WHO-Afro 2011). In Niger, where the national research institute has been using a similar protocol for more than 15 years, it has been estimated that 60% of health centres send CSF samples (Paireau et al. 2012) and 28–100% of reported cases are investigated (Collard et al. 2011). In other countries in the meningitis belt during 2011, the number of laboratory-investigated CSF samples represented 22% of reported suspected cases (WHO-Afro 2011). The reasons for incompleteness include the fact that in several countries, such as Mali and Togo, only doctors perform lumbar puncture, while health centres are staffed by nurses or medical assistants and even doctors regularly refrain from lumbar puncture. CSF transport from rural health centres to the district capital may not be funded or a burden for isolated health workers, and trans-isolate cultures are regularly contaminated [12% of CSF samples analysed in 2011 (WHO-Afro 2011)]. Difficulties in maintaining skilled personnel and storing material appropriately are thus barriers to appropriate surveillance in remote areas. In Burkina Faso, where the greatest financial and training effort has been made by the Meningitis Vaccine Project, 81% of CSF samples from suspected cases were analysed with serogrouping during 2011, corresponding to 61% of reported suspected cases (WHO-Afro 2011). At the height of the 2012 meningitis season, by week 14, a final result (bacterial aetiology and serogroup if meningococcal) was available for only 20% of suspected cases reported since week 1 (WHO-Afro 2012), while some serogroup-specific information was available for all districts in epidemic situation (data: Situation épidémique hebdomadaire, Burkina Faso Ministry of Health, 2012). Outbreak investigations in Burkina Faso have been supplemented during recent years by a mobile laboratory, which has improved timeliness (Betty Njanpop Lafourcade, personal communication).
The main limitation of Enhanced Meningitis Surveillance for assessing the impact of MenAfriVac® is the lack of comparable data before its introduction, as variation in protocols and resulting completeness make bias likely. The most adequate data are from Niger, where the same protocol with systematic PCR analysis has been used since late 2002 (Sidikou et al. 2003), 8 years before vaccine introduction. For epidemic detection and investigation, the criteria of timeliness and geographical coverage are not met in many settings with Enhanced Meningitis Surveillance; however, sufficient numbers of CSF samples appear to be collected in epidemic districts. For most research needs, the current Enhanced Surveillance protocol fulfils the essential criteria, where clinical and laboratory data are linked and where completeness is high. The system does not survey pneumonia and is currently of limited value for PCV impact assessment, as systematic information on the serotype of pneumococcal cases is reported only in Niger (Alio Sanda et al. 2011).
PCR-based surveillance systems
Several surveillance systems based on conventional PCR have been published previously (Massenet et al. 2009; Delrieu et al. 2011; Collard et al. 2011). They are part of the Enhanced Meningitis Surveillance system in most countries, but include systematic PCR analysis of CSF. Where completeness and geographical coverage are low, the resulting data have limited value for formal vaccine impact assessment, outbreak detection or research; nevertheless, PCR analysis has substantially increased the proportion of cases tested in the laboratory in remote areas, in countries with poor funding of meningitis surveillance and during investigation of epidemics, such as in Cameroon, Côte d’Ivoire and Mali. At other sites, such as the Burkina Faso national reference laboratory in Ouagadougou, real-time PCR is used. Both PCR techniques require only CSF collection in dry tubes at health centres and transport at ambient temperature with no strict delay. The management of skilled staff and stocks appears to be more sustainable in a surveillance system based on a central PCR laboratory than on decentralised culture and latex analyses (Chanteau et al. 2006). It remains to be seen which PCR technique is more appropriate in terms of diagnostic performance and infrastructural constraints. Conventional PCR can also include serotyping of pneumococcal cases directly on CSF, which is more suitable for the financial and infrastructural constraints than the classical Neufeld Quellung method for isolates (Njanpop Lafourcade et al. 2010).
Several meningitis surveillance studies involving active exhaustive case finding and systematic laboratory confirmation were published during the past decade. PCR was used systematically in studies in northern Ghana (Leimkugel et al. 2007), Burkina Faso (Siéet al. 2008; Delrieu et al. 2011), and northern Togo (Tall et al. 2012b), with source populations ranging from 140,000–865,000 inhabitants. Such studies are costly due to exhaustive laboratory diagnosis and quality assurance at all levels. While their contribution to outbreak surveillance is limited by low geographical coverage, they fulfil, if surveillance has started several years before vaccine introduction, criteria for vaccine impact assessment of MenAfriVac®. Only one site also reports pneumonia data in addition to meningitis (Tall et al. 2012b), but all conduct systematic serotyping and are therefore appropriate for PCV impact assessment.
Surveillance activities in reference hospitals are adequate for research as access to patients and quality laboratory analysis is facilitated, but their value for vaccine impact assessment depends on the representativeness of attending patients for the entire population and whether attending patterns are stable over time. For example, the Pediatric Bacterial Meningitis Network in paediatric wards of reference hospitals has been helpful for Hib vaccine impact assessment (Centers for Disease Control Prevention 2009), but is of limited value for surveillance of meningococcal and pneumococcal meningitis, which should also cover adolescents and adults.
Adapted surveillance strategy
To overcome these problems in surveillance, an adapted surveillance strategy for the meningitis belt is proposed that takes into account infrastructural and financial limitations. To assess MenAfriVac® impact, current surveillance activities that have historical data and good completeness should be maintained for up to 10 years after vaccine introduction, such as in PCR-based Enhanced Meningitis Surveillance in Niger (Sidikou et al. 2003; Collard et al. 2011) and a surveillance study in northern Ghana (Leimkugel et al. 2007)). Furthermore, surveillance studies that were recently discontinued might be resumed, such as the surveillance study in western Burkina Faso (Parent du Chatelet et al. 2005; Delrieu et al. 2011)). In all countries, weekly data on suspected cases should be compiled and analysed in health centre level resolution to assess trends in the annual incidences of localised epidemics. Vaccine effectiveness can be evaluated anywhere that NmA incidence resumes and valid surveillance can be instituted. To prepare future PCV impact assessment, Enhanced Meningitis Surveillance systems with wide geographical coverage and good completeness (in Burkina Faso and Niger) should introduce systematic serotyping of pneumococcal cases, for example by PCR on CSF.
For more timely detection of epidemics, the existing reporting system of suspected cases should include weekly analysis of data in health centre level resolution in all countries, and apply, e.g. the recommended epidemic threshold for districts with fewer than 30 000 inhabitants (Tall et al. 2012a; Paireau et al. 2012). Furthermore, systematic lumbar puncture should be abandoned for surveillance purpose, as lumbar puncture frequently has side effects and should be reserved for situations in which it can effectively guide clinical decision-making, or when it is part of a scientific protocol for vaccine impact assessment or other research requiring laboratory confirmation. For timely identification of epidemic agents, mobile outbreak investigation teams should initiate systematic lumbar puncture and laboratory diagnosis as soon as a signal is detected at a health centre, including CSF transport to the nearest laboratory for PCR, culture, or rapid tests. In the context of limited infrastructure and funding, only a simplified protocol such as that outlined above may be feasible and cost-efficient.
In the long term, surveillance studies should be maintained throughout the meningitis belt, each study covering at least 0.5 million inhabitants with exhaustive case identification and high-performance diagnostic tools. Ideally, the sites would also undertake surveillance of pneumococcal pneumonia or septicaemia and survey other pathogens. In addition, data on suspected case reports should be compiled and analysed in health centre level resolution, to allow research on risk factors for meningitis epidemics.
Surveillance systems and techniques for diagnosing bacterial meningitis in the African meningitis belt have improved greatly over the past few decades, but current surveillance strategies are insufficient in many settings for impact assessment of new vaccines and for identifying epidemic meningococcal serogroups in a timely manner. Vaccine development should be accompanied with long-term funding of relevant surveillance systems, and current and new surveillance techniques should be evaluated for their sustainability and cost-effectiveness. For both these objectives, better coordination is needed between individual research groups, countries, international organisations and funders.
The author thanks Stéphane Hugonnet and the peer reviewers for helpful criticism of the manuscript.