Electronic data in support of new drug applications have been the standard for submittal to regulatory authorities – ever since computers were found to be reliable and their prevalence made them readily available. For much of that time, however, the source documents on which the clinical data was initially recorded were paper. These were typically paper medical charts and paper CRFs. This required manual transcription from the paper source documents into the computer system as electronic records and was typically handled by data entry personnel working at the investigative site.
The extra step for transcription required resources to perform the work – incurring additional costs. This extra work also took time to complete. Given typical patient enrollment at a site and the dispersed timing of visit schedules for those patients, the data entry step was only a part-
time task. Sites tended to wait for data to accumulate and then doing data entry on the bolus of source documents. This delay meant that the sponsor did not have the data for the earlier visits in a timely fashion, which had repercussions in monitoring study results, potential adverse events, and protocol adherence by the site.
In addition to the time and costs to complete this transcription step at the site, the process introduced a source of error due to the risk typographical and other errors in the EDC or study database. In order to identify and correct these errors, a process called ‘source data verification‘ (SDV) was assigned to CRAs who monitored the investigative site. Additional time and expense were incurred so that CRAs could visit sites to accomplish this manual task. Any errors that the CRA identified initiated a data query, which was a request by the CRA to the site to review and correct questionable data. So, additional time and expense were expended on the correction process.
Research conducted over the past five years or so has confirmed that SDV has minimal benefit. Most of the errors identified in the process were for data points that had no significant impact on trial endpoints. Clinically significant data changes were identified in 1% to 2% of all data reviewed. This called into question the benefit of SDV relative to the cost of on-site monitoring visits.
Based on this information, many data management teams started to identify which data points to be collected in a study that were clinically significant. SDV was then focused on comparisons of those data between the source documents and the EDC. This limited SDV reduce the overhead for on-site monitoring, as well as focused attention on key data points, making the process more effective.
That said, SDV was still considered questionable from a cost-benefit perspective. However, sponsors and CROs were hesitant to de-emphasize it, lest they incur regulatory issues down the road. For this reason, even at this late date, SDV continues to play a significant part in the site monitoring story.
From a data quality perspective, the ideal process would have patient data input directly into the clinical study database (along with all relevant metadata). If that data could be created and saved via an automated process, removing human intervention, that would be even better. This is not the case, however. Because clinical care of the patient is foremost, and supersedes the needs a clinical trial, most patient data is captured in an Electronic Medical Record (EMR) system. Some of this data is generated and captured automatically – vitals signs via devices that are attached to the EMR infrastructure, for example. However, some data is generated via HCP observations and manually recording in the EMR, which introduces the risk of error via typographical mistakes.
Once the patient data is captured in the EMR, some of it is required by the study protocol and so those datapoints must be copied to the study database – either directly or via an intermediary system, such as EDC. The currently predominate process is manual transfer into the EDC – essentially a data entry step. It is this step that incurs source of errors risk prompting the SDV monitoring step. If the EMR data could be copied programmatically into the study database via a validated process, we would effectively remove the risk of errors between EMR and study database. We would also remove the need for costly and time-consuming monitoring steps.
Next up: Managing source data; Data Quality; Source Document Verification; EHR; and Decentralized Clinical Trials