A feasibility assessment of real‐world data capabilities for monitoring vaccine safety and effectiveness in China: Human papillomavirus vaccination in the Yinzhou district as a use case

Real‐world data (RWD) are increasingly used to generate real‐world evidence (RWE) of vaccine safety and effectiveness for regulatory purposes. Assessing feasibility of using RWD sources prior to implementing observational studies is recommended. As a use case, we described the process and findings of a feasibility assessment to identify reliable and relevant data sources for monitoring the safety and effectiveness of the AS04‐HPV‐16/18 human papillomavirus (HPV) vaccine in China.


Key Points
• Use of RWD to generate RWE is an integral part of post-authorization vaccine monitoring.
• Feasibility assessment of RWD sources prior to their use supports good pharmacoepidemiological practice and is increasingly recommended by regulatory authorities.
• HPV vaccines are presented as a use case to assess the feasibility of the YRHIP as an appropriate RWD source for the evaluation of vaccine safety and effectiveness in China.
• These findings confirm the appropriateness of the YRHIP for this purpose.
• This is supported by a growing number of studies reporting on YRHIPs capabilities to generate RWE.

Plain Language Summary
Data from the use of medicines in large populations after their approval are called RWD. RWE generated from these data are increasingly used to inform the safety and effectiveness of vaccines. Systems combining RWD from different sources such as electronic health records of patients or disease and vaccination registries are used to generate the comprehensive information requested by regulatory agencies for the continuous evaluation of the benefit/risk of vaccines. Following approval in China of the AS04-HPV-16/18 vaccine, which protects against HPV genotype 16/18 infection and related cancers, there was a need to identify appropriate RWD sources to monitor its effects in routine use. The feasibility of using the YRHIP was evaluated and found that it provides all necessary information to generate RWE on HPV vaccine safety and effectiveness. This is supported by a growing number of studies using this platform for a broad range of diseases and other vaccine studies. Studies using the YRHIP to monitor the HPV vaccine in routine use building on these findings are underway.

| INTRODUCTION
Identification and analysis of real-world data (RWD) to generate realworld evidence (RWE) is increasingly used to inform regulatory decision-making, 1 as well as payer and reimbursement policies on medicinal products and devices. [2][3][4][5] Use of RWE is established for post-authorization monitoring of medicinal products by regulatory authorities in North America and Europe, [6][7][8][9][10][11] and increasingly elsewhere, including the National Medical Products Administration (NMPA) in China. 12 While randomized controlled trials (RCTs) remain the gold standard for approval, RWE is used in post-authorization safety and effectiveness studies, 2,13-15 although there is a shift toward RWE acceptance pre-and peri-licensure. 7,16 Criteria for RWE acceptability are selection of RWD sources, reliability, and relevance of the data 1,17,18 and ability to link exposure (immunization) data to health outcomes at individual level. It is considered good pharmacoepidemiological practice to conduct a feasibility assessment of RWD sources prior to their use. 14,15,19 Although use of RWD and acceptability of RWE in China is at an earlier stage compared to other countries, RWD capabilities and RWE acceptability are developing rapidly, 20,21 including the capability to monitor adverse events following immunization (AEFIs). 22,23 The National Medical Products Administration (NMPA) has published corresponding guidance. 24,25 As elsewhere, substantial advances in healthcare record digitization have been made in China. 26 A wide array of systems now exist at local, regional and national levels, spanning electronic health records (EHRs), health insurance databases, disease and birth/death registries, providing a sound foundation for real-world studies. 21,26,27 The bivalent AS04-adjuvanted HPV-16/18 (AS04-HPV-  [28][29][30][31][32] At the time of the work presented in this article, the indication was for use in women 9-45 years of age. While a favorable safety profile was observed during global clinical development, 33 as a condition of approval the NMPA requested evaluation of vaccine safety and effectiveness using appropriate high-quality RWD sources.
In line with good pharmacoepidemiological practice and international standards, the purpose of this article is to describe the stepwise feasibility assessment that led to the identification of a fit-for-purpose RWD source for the evaluation of vaccine safety and effectiveness in China, using monitoring of the HPV vaccine safety and effectiveness as a use case that can be extrapolated to other vaccine studies. This choice was justified by the availability of this vaccine since mid-2017, in the context of adult immunization which is not systematically electronically recorded in China. This exercise, initiated in response to post-authorization requirements following initial NMPA approval was conducted at a time when information on the suitability of RWD sources to support real-world vaccine monitoring in China was limited.

| METHODS
A multi-step iterative approach was adopted, involving input from a wide range of information sources, as shown in Figure 1.
Step 1. Literature review and data source mapping PubMed and Embase were queried to retrieve literature about studies using claims, EHRs and registry data (vaccination and/or cancer) in China. The aim of this targeted review was to identify potentially relevant RWD sources while in parallel mapping cervical cancer screening and adult vaccination rates.
Step 2. Expert opinion Insights from national experts in RWD during the 11th Asian Conference on Pharmacoepidemiology (ACPE) in Xi'an, China, in October 2018 was sought to collect knowledge and field experience on the availability and use of RWD sources to complement our targeted literature review.
Step 3. Survey To ensure the selected RWD source was able to generate meaningful RWE, a survey was conducted based on published methods 15,20,34 to assess the data in terms of population characteristics (size, coverage, representativeness), data lag, data refresh, data linkage, as well as exposure (immunization data including brand, dosage and vaccination dates), outcome and potential confounding variables for our use case. Safety outcomes included, amongst others, potential immune-mediated disorders (pIMDs), including type 1 diabetes mellitus (T1DM), autoimmune thyroiditis (AIT), and inflammatory bowel disease (IBD). [35][36][37] Effectiveness outcomes included data from cervical cancer screening results. The survey was administered to experts from Beijing University, Health Science Center, given their extensive experience in using the YRHIP.
Step 4. Literature monitoring Continued surveillance of the published literature supported contextualization of findings of studies using the selected source of RWD, as many studies were in progress at the initiation of this feasibility assessment and only published at a later stage of writing of our results.

| RESULTS
Key findings of this feasibility assessment are described below.
Step 1. Literature review and data source mapping Province compared with other regions, with well-reported historical epidemiological data. [39][40][41] Furthermore, adult vaccination uptake appeared high in this area. 42,43 Step 2. Expert opinion ACPE in Xi'an provided an important opportunity to leverage insights on the suitability of various RWD sources for vaccine monitoring across the Asia-Pacific region. 44 Discussions organized during the conference between the authors and governmental, academic and industry experts enabled to better understand the RWD landscape and its challenges and opportunities.
F I G U R E 1 Feasibility assessment approach. CDC, Center for Disease Control and Prevention; HPV, human papillomavirus; RWD, real-world data; YRHIP, Yinzhou Regional Health Information Platform.
Preliminary findings from Steps 1 and 2, combined with the previous evaluation RWD sources in China by Yang et al. 20 confirmed that the YRHIP was fit-for-purpose for our use case.
Step 3. Survey Results of the survey described hereafter, combined with findings from Steps 1 and 2, showed that the YRHIP population size and coverage, as well as exposure, confounding and outcome data, including pIMDs for safety, and HPV-related outcomes for effectiveness were adequately recorded and linkable at individual level across the different databases in the platform.
The YRHIP is a unique integrated electronic health information (ICD-10) codes. 38 Importantly, 98% of all permanent residents (more than 1.2 million individuals) with de-identifiable data linkable using a unique key common across databases (including electronic health records (EHRs) and immunization registries) were already included in 2018. This allows linkage to a range of EHRs and disease, mortality and immunization registries as well as health insurance and pharmacovigilance reporting databases, creating a fully integrated platform, with data updated in near real-time. 38 The YRHIP also contributes to a broader Regional Health Care Information Platform for Ningbo City, which in turn has the capacity to contribute to supra-regional data platforms. This illustrates how integrated data platforms co-exist at different scales in China (i.e., district within a city and the larger city) as each platform has the capacity to access integrated data for unique individuals on the basis of their geographical residence.
Our assessment of the YRHIP confirmed the capability to capture immunization, safety and effectiveness outcome data, from which meaningful RWE can be generated. Using HPV vaccine monitoring as a use case, this work specifically evaluated the operational infrastructure of the YHRIP, evaluating the system's capacity to record and link core variables for all individuals with accessible records, that is, subject demographics, medical information from EHRs, maternal and child health data, immunization registration and AEFI endpoints. Ability to identify individuals eligible for HPV vaccination as well as key safety and effectiveness outcomes via ICD-10 coding was also confirmed.
This work highlighted that, as of end of 2018, there were an estimated 250 000 permanent female residents aged 9-45 years (and eligible to receive HPV immunization) registered within the YRHIP. In addition, information was available on pregnancy outcomes (e.g., still births, spontaneous abortions, preterm delivery, congenital malformations) for 110 000 females across 2012-2018. All these represent a F I G U R E 2 Overview of databases integrated within YRHIP. *Immunization data includes administration date, vaccine product name, manufacturer, batch No., dosage, and number of vaccinations. ** Detailed demographics are also recorded in Primary care databases. AEFIs, adverse events following immunization; CDC, Center for Disease Control and Prevention; EHR, electronic health records; ICD-10, International Classification of Diseases, Tenth Revision; YRHIP, Yinzhou Regional Health Information Platform.
T A B L E 1 Real world data studies utilizing Yinzhou Regional Health Information Platform. Abbreviations: AEFI, adverse events following immunization; CKD, chronic kidney disease; CVD, cardiovascular disease; HPV, human papillomavirus; HZ, herpes zoster; ICD-10, International Classification of Diseases, Tenth Revision; pIMDs, potential immune mediated disorders; RA, rheumatoid arthritis; RWD, real-world data; YRHIP, Yinzhou Regional Health Information Platform.
Step 4. Literature monitoring Evidence from continuous updates of our literature review corroborated these results. More recently published evidence points to the importance of identification of HPV16/18 genotyping data as baseline data for the period prior to vaccine introduction to assess type-specific effectiveness using YRHIP data. 46 The suitability of the YRHIP for RWE generation in China is increasingly recognized. In their appraisal of the characteristics of different electronic data sources for use in post-marketing safety surveillance, Yang and colleagues reported that the YRHIP integrated a far greater array of relevant variables than alternative EHR databases. 20 A number of studies in progress at the time of this assessment are now published, confirming the capability to generate RWE from large populations in the YRHIP (Table 1). 38 vaccine, highlighting the capacity to differentiate vaccine products.
All relevant variables were available for all identified vaccine recipients. The capability to identify specific pIMDs occurring within 30-or 90-days following vaccination and generate pIMD incidence rates was also confirmed. 53 Although reporting across a broader population, these observations confirm the capacity to integrate anonymized subject-level data from different sources (EHRs, maternal health and immunization registries) to identify pIMDs in women receiving a HPV vaccine, supporting the YRHIP's suitability for vaccine monitoring.

| DISCUSSION
This iterative feasibility assessment supports the use of the YRHIP as a suitable RWD source for post-authorization vaccine safety and effectiveness studies, as suggested by our use case, which showed that data recording and integration allows identification and analysis of all necessary information to conduct HPV vaccine surveillance in the Yinzhou population. This was further supported by subsequently published studies (in progress at the time of our assessment) reporting on the capacity to record a broad range of outcomes in the large resident population captured in the YRHIP. 38,45,[47][48][49][50][51] These include studies evaluating vaccine uptake and safety, either as a discrete platform, 45,56 or when contributing to the broader Ningbo City Regional Health Care Information Platform. [52][53][54][55] However, there is some uncertainty to which extent more recent and future data may In both studies, this probabilistic approach will mitigate limitations due to uncertainty on sensitivity/specificity of endpoint identification, baseline pIMD incidence rates, cervical lesion prevalence rates, HPV vaccine uptake and required sample size for study design. These studies will provide a greater understanding of the validity (sensitivity, specificity and predictive values) of case definitions and case finding algorithms in the YRHIP and precision of derived data.
While our use case highlights the strengths of the YRHIP for HPV vaccine monitoring, it reflects outcomes in one, highly developed, area in the southeast of China, which may reduce generalizability of findings, both for our use case and for extrapolating to other vaccines.
Nevertheless, from a targeted literature mapping, combined to expert opinion and an independent survey, our approach enabled a comprehensive assessment of existing RWD sources in China, while focusing on a selected platform as most suitable to address our use case.

| CONCLUSIONS
Our use case illustrates the value of assessing feasibility of potential RWD sources for vaccine monitoring prior to RWE safety and effectiveness studies. 15 The identified suitability of the YRHIP is supported by studies subsequently reporting across a range of populations and objectives. 38,45,[47][48][49][50][51] While further validation of HPV vaccine-related endpoints is ongoing, our work suggests that data characterization prior to any study helps mitigate some limitations of the YRHIP and tailor