Secondary Logo

Journal Logo

Special Article

Progress Toward a Global Vaccine Data Network

Petousis-Harris, Helen PhD*; Dodd, Caitlin N. PhD

Author Information
The Pediatric Infectious Disease Journal: November 2020 - Volume 39 - Issue 11 - p 1023-1025
doi: 10.1097/INF.0000000000002785
  • Free

The continued success of the global vaccination effort relies upon both the use of safe vaccines and public confidence in their benefit and safety. In recent years, there have been examples of dramatic falls in the gains made in vaccine coverage and disease prevention in both the developed world and in low and middle-income countries, where scares have resulted in loss of public confidence. While vaccine hesitancy and antivaccine communication has become global, the ability to respond to such concerns has remained largely fractured without coordination between countries.

Assessment of adverse events following immunization (AEFI) that have a delayed onset or diagnosis, occur beyond clinical trial study follow-up, are rare or occur among subpopulations are often beyond the scope of initial clinical programs. This makes postlicensure studies critical. In the past decade, international collaborations have demonstrated the ability to supply background rates of events of interest in anticipation of the use of new vaccines such as H1N1 pandemic vaccines, develop tools and procedures for collaborative studies of vaccine safety and efficacy, and demonstrate capacity to conduct vaccine safety studies globally.1

Isolated collaborations that have included multiple countries for active surveillance using pooled data have generated information on measles-containing vaccines and aseptic meningitis and idiopathic thrombocytopenic purpura,2,3 rotavirus vaccines and intussusception4 as well as pandemic influenza vaccine and Guillain-Barre syndrome and narcolepsy.5 Sustainable infrastructure and capacity building are needed for both timely hypothesis testing and to assess new vaccines such as Ebola, dengue, malaria and typhoid introduced particularly into LMICs.

Clearly, such collaborations are feasible; however, they also need to be sustainable, responsive and cost-effective.

Given the unprecedented rate of vaccine development in response to the COVID-19 pandemic and the potential for a diversity of vaccine formulations, manufacturers and strategies, a prepared global network to monitor the safety and effectiveness of these vaccines will be needed. It will be of critical importance to monitor the safety of these vaccines during their deployment, preferably in near real time, for large and diverse populations. This is also going to be vital if trust in vaccines is to be maintained through the pandemic period.

Based on the progress and evaluation of the Global Vaccine Safety Initiative Blueprint 1.0, strategies for the next decade have been highlighted in Blueprint 2.0.6,7 These include the measurement of background rates for events of interest in active surveillance and sentinel systems and the coordination existing in active surveillance and sentinel systems nationally, regionally and globally to take advantage of variability in vaccination practices and to increase power and timeliness.

The Global Vaccine Data Network (GVDN) was established in 2019 as a network of over 16 sites in 14 countries1 and 5 WHO regions who have demonstrated capacity in vaccine safety studies using clinical databases either independently or as collaborators. Between these sites, there is an estimated population of over 235 million under surveillance. New partners representing additional sites in low- and middle-income countries in Africa and South Asia have been identified and approached based on their potential to contribute through capacity building. The consortium is supported by a coordinating center based at the University of Auckland in New Zealand.

The GVDN has been established for conducting international studies of vaccine safety, efficacy and risk–benefit through a governance model that ensures full participation of sites on a voluntary basis from study conception through to study reporting. The network is governed by a geographically diverse executive committee of seven site leaders and two (co) directors. Coordination including functions such as process, finance, secretariat and data management is provided by the coordinating center (Table 1).

TABLE 1. - Responsibilities of the Coordinating Center
Science ∘ Support consistency of research outputs
∘ Support manuscript development
∘ Coordinate advice and mentoring
Process ∘ Coordination of global studies and proposals
∘ Establishment and maintenance of terms
∘ Active management of live studies
Financial ∘ Fundraising and business development
∘ Financial management of projects
Services ∘ Contracts management
∘ Legal council
∘ PR/external communications
∘ Secretariat
∘ Website—development and maintenance
Tools ∘ Secure web-based platform for data and analysis
∘ Common data model(s)
∘ Modular, reusable data transformation and analysis scripts

To protect privacy but allow collaborative agreements, the network will use common data models (CDMs) for the participating sites that are flexible in terms of allowing for diverse data sources. This also allows for tiered participation dependent upon data availability while allowing for participation in the network as capacity is developed in a site. Data analysis models are defined that protect data privacy while also allowing for collaborative analysis of data. With local data in a CDM, centrally written scripts can be deployed against local data to produce individual-level analysis sets to be pooled centrally, aggregated minimum analysis sets to be analyzed centrally or model estimates to be meta-analyzed centrally. It is also possible to utilize a hybrid of these approaches dependent upon capacity and data protection needs of each site. A framework for determining the analysis model to be used for a given study will be dependent upon study question(s) and participating network sites. To this end, the network has a secure platform for data sharing and collaborative analysis.

An assessment of site capacity and gaps is presently being addressed via the development of a matrix of potential useful data and databases for the network; definition of site characteristics and capacities to include in the matrix; the mapping of each site, including experience, data sources and characteristics to the matrix; an assessment of which capacities and data are necessary for each type of study (cohort, case control, self-control case series, background rates) and identification of key gaps by importance and frequency. This exercise is also enabling the assessment of the ability of the sites to work collaboratively.

To be a participating site, the following conditions need to be met:

  • The site has an investigator willing to participate in the network and willing to review potential study evaluations to determine if they are willing and able to participate.
  • Demonstrated capacity to perform studies of vaccine safety and epidemiology.
  • Ability to identify potential vaccine safety outcomes of interest through hospital databases or demographic surveys and ideally outpatient data bases as well.
  • Ability to assess vaccine exposures either electronically or through review of medical records

However, it must be noted that not all sites have equal capacity to engage in all types of studies. Many are using different methods to ascertain the same information. Some sites have a population denominator whereas others use self-control case series methods to obviate the need for such a denominator. Some sites have comprehensive immunization registries, whereas a minority of sites require accessing paper records to obtain this information.

CAPACITY BUILDING

Another goal of the GVDN is to assess and develop capacity in LMIC sites which currently have not demonstrated capability and likely require capacity development to participate in the network. This component is critical to the future of vaccine safety assessment of new vaccines which are being and will be introduced either exclusively or almost exclusively in LMIC, such as malaria, typhoid, shigella and tuberculosis vaccines. We have identified sites which show potential for capacity building and mentoring into a collaborative model and propose to conduct a validation exercise regarding the ability to identify and capture key data elements, perform an analysis of gaps in capacity, and make recommendations as to how to address those gaps in capacity.

COORDINATION

The GVDN requires a range of operational support functions. Financially, there is a need for a funding framework for projects, and accounting to multicountry sites, in multiple currencies. Legal requirements include terms of references for partners, assurance of data privacy and appropriate treatment of personal information. Finally, administration functions that include contracts management with sponsors and sites, business development and a secretariat to liaise with network members.

COMMON DATA MODEL

The GVDN proposes data models that protect individual privacy but allow collaborative agreements on CDMs (standardization of data to allow pooling of results).

The use of CDMs to facilitate generation of evidence using real world data in collaborative networks has seen a marked increase in popularity. CDMs harmonize the structure and/or content of the participating data sources to enable the use of common analytics developed centrally to run against all data sources. Recent collaborative vaccine safety studies have made use of study-specific CDMs in which data held by contributing partners are mapped to a set of concepts relevant to the study question. While this has been effective to allow for use of common analysis tools across sites, it does not increase data readiness to address a variety of study questions. Other more flexible approaches include those employed by sustained networks such as the US Vaccine Safety Datalink (VSD),8,9 the Observational Medical Outcomes Partnership (OMOP)10 and the US FDA Sentinel System.11 Both the VSD and OMOP use a mapping CDM, in which diverse underlying data sources are mapped to a set of common concepts using a set of common terminologies such as Anatomical Therapeutic Chemical (ATC) Classification System codes for drugs and vaccines. The Sentinel CDM, alternatively, has elements both of a mapping CDM and an organizing CDM. In an organizing CDM, diverse underlying data sources are mapped to a common structure, but the content remains in its original format.12

We are developing a CDM for the participating sites which is flexible enough to include diverse data sources such as health and demographic surveillance systems, cohorts, electronic health records, hospital records and population statistics. The CDM will contain tables to accommodate data on populations, observation periods, death, exposure, events, procedures, mother–child linkage as well as others should their need become evident. Tables may be empty if data are not available. We foresee that this model will allow for tiered participation dependent upon data availability, while allowing for participation in the network as capacity is developed in a site.

To formalize this CDM, we will review existing CDMs for their utility for conducting vaccine safety studies, assess data availability at each study site, review data dictionaries of participating sites and update the CDM in an iterative manner until consensus is reached. This consensus building process will include decision making regarding to what extent we choose to employ a mapping CDM (as in the VSD), an organizing CDM or a hybrid of the two (as in Sentinel).

MODELS FOR ANALYSIS

In addition to formalizing a CDM for the network, we will define models for data analysis which protect data privacy while also allowing for collaborative analysis of data. Potential analysis models vary in how much individual-level data is shared and how data or results are pooled. With local data in a CDM, centrally written scripts can be deployed against local data to produce individual-level analysis data sets to be pooled centrally, aggregated minimum analysis sets to be analyzed centrally or model estimates to be meta-analyzed centrally. It is also possible to utilize a hybrid of these approaches dependent upon capacity and the data protection needs of each site. We will assess these potential solutions through review of existing networks, recent studies on the performance of privacy protecting analysis methods1,2,13 as well as the needs and capacity of participating sites.

MOVING FORWARD

A proposal to utilize this infrastructure to evaluate the association with GBS with vaccination including an evaluation of potential genetic risk factors has been submitted for funding to NIH. In addition, in February 2020, the network has begun discussions as to how the network can address the need for a global comparative study of COVID-19 vaccines when these vaccines become available.

In conclusion, capacity currently exists globally to assess vaccine safety across a network of geographically diverse sites. Developing this capacity further will be critical to the assessment of COVID-19 vaccines and other newly introduced vaccines in the future.

REFERENCES

1. Black S, Eskola J, Siegrist CA, et al. Importance of background rates of disease in assessment of vaccine safety during mass immunisation with pandemic H1N1 influenza vaccines. Lancet. 2009; 374:2115–2122
2. Guillard-Maure C, Elango V, Black S, et al.; WHO Global Vaccine Safety-Multi Country Collaboration. Operational lessons learned in conducting a multi-country collaboration for vaccine safety signal verification and hypothesis testing: The global vaccine safety multi country collaboration initiative. Vaccine. 2018; 36:355–362
3. Perez-Vilar S, Weibel D, Sturkenboom M, et al.; WHO Global Vaccine Safety-Multi Country Collaboration. Enhancing global vaccine pharmacovigilance: proof-of-concept study on aseptic meningitis and immune thrombocytopenic purpura following measles-mumps containing vaccination. Vaccine. 2018; 36:347–354
4. Tate JE, Mwenda JM, Armah G, et al.; African Intussusception Surveillance Network. Evaluation of intussusception after monovalent rotavirus vaccination in Africa. N Engl J Med. 2018; 378:1521–1528
5. Weibel D, Sturkenboom M, Black S, et al. Narcolepsy and adjuvanted pandemic influenza A (H1N1) 2009 vaccines—multi-country assessment. Vaccine. 2018; 36:6202–6211
6. World Health Organization. Department of Immunization VaB. In: Global Vaccine Safety Blueprint. 2012. Geneva: Quality, Safety and Standards unit of the Department of Immunization, Vaccines and Biologicals
7. WHO Global Vaccine Safety Blueprint (GVSB) 2.0 Drafting Group. Global Vaccine Safety Blueprint 2.0Geneva: World Health Organization; DRAFT for Public Consultation
8. Chen RT, Glasser JW, Rhodes PH, et al. Vaccine safety datalink project: a new tool for improving vaccine safety monitoring in the United States. The vaccine safety datalink team. Pediatrics. 1997; 99:765–773
9. McNeil MM, Gee J, Weintraub ES, et al. The vaccine safety datalink: successes and challenges monitoring vaccine safety. Vaccine. 2014; 32:5390–5398
10. Observations Health Data Sciences and Informatics. 2020. Available at: https://www.ohdsi.org/data-standardization/the-common-data-model/. Accessed May 7, 2020
11. Ball R, Robb M, Anderson S, et al. The FDA’s sentinel initiative—a comprehensive approach to medical product surveillance. Clin Pharmacol Ther. 2016; 99:265–268
12. Schneeweiss S, Brown JS, Bate A, et al. Choosing among common data models for real-world data analyses fit for making decisions about the effectiveness of medical products. Clin Pharmacol Ther. 2020; 107:827–833
13. Shu D, Yoshida K, Fireman BH, et al. Inverse probability weighted Cox model in multi-site studies without sharing individual-level data. Stat Methods Med Res. 20191–14
Keywords:

Vaccine safety; distributed data networks; common data models

Copyright © 2020 Wolters Kluwer Health, Inc. All rights reserved.