Other pressures on researchers prevent them from sharing data. For example, results that run counter to the expectations of the sponsor or funder or the possibility of future publications may lead a researcher to want to keep the data private. Data ownership could also be seen as a grey area that could prevent the introduction of the data-sharing mechanism. The privacy of personal data in the UK is covered by the common law and the Data Protection Act (1998) [2], which follows the EU Data Protection Directive (1995). Data protection laws exist in countries outside the EU with similar protection. The common law and data protection law of the United Kingdom covers identifiable data about living persons and data that could be identified from other information that the controller has or is likely to receive. If you have your explicit consent to the disclosure of personal data, you may share it (as indicated in the consent). If the participants have died, they are not subject to the Data Protection Act. Otherwise, data can only be shared if it is anonymized and the data is not identifiable – and identifiability must take into account the identification that could be provided by the information that the controller has or is likely to receive. Solving these problems takes time, which can also prevent researchers from sharing their data. It may be necessary to establish a formal data-sharing committee as well as a concerted effort to promote data availability. Appropriate documentation should be integrated into formal processes for all UTCs (e.g.B. via Standard Operating Procedures (SOPs)).
This future work will also have an impact on resources, but should lead to a positive step forward in the area of data exchange. Mowat C, Arnott I, Cahill A, for the TOPPIC study group, et al. Mercaptopurine versus placebo to prevent recurrence of Crohn`s disease after surgical resection (TOPPIC): a multicenter, double-blind, randomized controlled trial. Lancet Gastroenterol Hepatol. 2016;1:273–82. Other studies that go through the same process within the ECTU are BIDS [16] and GaPP [17], both of which have been published and have provided valuable information in the data anonymization process, which shows in particular that a standardized solution is not always the most appropriate. The data will be used to develop an PGD REPOSITORY on potential facilitators, health outcomes, use and cost of healthcare resources, from RCTs that test therapist-led interventions for back pain. We will perform statistical and health economic analyses on this aggregated dataset. The original data dictionary serves as a starting point for documenting the anonymization process, assigning a value to each variable. Direct identifiers receive values 01 through 14, indirect identifiers receive values A to N, and superfluous information receive values 15 (Table 1).
Data come from completed randomised controlled trials. All data is anonymous and no information that can identify the patient is shared. For TOPPIC, one serious adverse event (SAE) related to the child of a patient born during the TOPPIC follow-up period was recorded. Several feasible solutions were presented: (1) anonymization of the SAE dataset but link to toppic patient (parent), (2) anonymization of the SAE dataset and destruction of the link to TOPPIC patient, and (3) complete deletion of data. It was decided that this CAS record should be anonymized, but the patient link to an indicator should be retained to indicate that the CAS was related to the child (Solution 1). This way, people would remain unidentified, but this rare event would be stored in the anonymized data. Your research proposals will be reviewed by a committee of internal consultants. For clinical trials that are subject to agreements with co-development partners, Amgen will work with relevant partners on data exchange requests.
In general, Amgen does not support external research issues that include access to individual data at the patient level to reassess safety and efficacy issues that have already been addressed in product labeling. If the result of the internal review is to reject the application, an Independent Data Sharing Review Committee (IRB) will arbitrate and make the final decision. The following people make up the DSIRP: A cheaper alternative to a full repetition of the original analyses would be to check some important facts of the main analysis and use simple automated checks for the rest. When programming the anonymized version of the record, it is more likely that an incorrect variable will be used instead of the variable that only one point of the record will be changed. Therefore, the comparison of the difference between the maximum and minimum values in anonymized and original datasets is likely to be sufficient for variables and continuous data. The folder in which this data is stored must be protected so that it cannot be unintentionally modified. The production of data dictionaries can be automated, but for the TOPPIC study it was a retrospective and relatively manual process. Once the data anonymization process was complete, the recording was exported to a comma-separated text format (i.e., comma-separated variables (csv)). This can be easily read in widely used packages such as Microsoft Excel. It has the advantage of being readable for a long time in the future, while other data tables are not readable because software versions are updated over time. However, an exception to this rule would be if the values contain commas (for example.B. if there are text fields).
Clinical Study Data Request (CSDR) is a consortium of 13 international pharmaceutical companies (GSK, Astellas Pharma, Bayer, Chugai, Eisai, Novartis, ONO, Roche, Sanofi, Sunovion, Shionogi Inc, UCB and ViiV) and four academic research funders (The Wellcome Trust, The Bill & Melinda Gates Foundation, The UK Medical Research Council and Cancer Research UK).1 It was launched in 2013 and currently lists anonymized patient data from 3374 studies on the platform. including 10 studies conducted by university funders. The mandate is to remove barriers to access to and re-use of data, thereby facilitating the exchange of data in a fair, transparent and independent manner. A unique identifier for each study participant will always exist, and very often the assigned unique identifier of origin can be linked to the study sites. Therefore, all unique identification numbers (p. ex. B, subject number, pre-selection identifier) were recoded using the random number generator method, which ensured reproducibility and association with the original unique identifier (see Supplementary File 1). The MRC-HTMR guidelines suggest removing the link between the new code number and the original unique identifier.
For the TOPPIC study, the link was kept for secondary researchers in case of questions about the anonymized dataset. Pressure to share data can be seen as both positive and negative, but should be underpinned by the need to ensure patient trust at all times [3, 4]. Trying to balance this pressure leaves researchers trapped in the middle. Guidelines on data exchange are available [5,6,7,8,9,10] and there is growing interest in determining the best methodology to carry out these processes [11, 12]. For researchers like us who work in university study units in the UK, the MRC HTMR guidelines [5] are particularly useful. The MRC HTMR guidelines state that at the end of a study, trialists must prepare a set of anonymized data ready for sharing after establishing an appropriate level of anonymization. The creation of the dataset should be carried out by individuals with an understanding of data management and basic statistics, and there should be independent quality control. The recording must be in a form recognized by a set of software. The sharing package should include supporting documentation, including the protocol and annotated data collection forms (including any changes throughout the study).
While much has been written, there are still gaps in the detail of what researchers need to do to share data securely. In particular, the anonymization process is not described in sufficient detail. The second passage of the data dictionary includes indirect identifiers, those that may pose a risk if present in combination with others. To decide whether these should be anonymized, a consensus model consisting of a study manager, a statistician and a computer programmer was used. .