Risk-Based Source Data Verification Approaches: Pros and Cons

Abstract

The high cost of source data verification (SDV), particularly in large trials, has made it a target of scrutiny over the last decade. In addition, the positive impact (ie, cost-benefit ratio of SDV) on overall data quality is often questioned. As a result, regulators and industry groups have started looking at alternative SDV approaches. This article evaluates the FDA-supported risk-based approach to SDV and provides a proposal on how to modify the SDV process without undermining the validity and integrity of the trial data. It summarizes alternative approaches to 100% SDV and evaluates the advantages and disadvantages of risk-based SDV (rSDV). The regulatory, data quality, and cost implications of each approach are considered. The economics of rSDV are discussed and the cost implications of rSDV are presented based on the results of exploratory analyses for four hypothetical trials in cardiology and oncology.

Efficient Source Data Verification Using Statistical Acceptance Sampling: A Simulation Study

“…Other approaches that have been proposed are, for example, ''targeted SDV'' (ie, focus SDV on high-risk data points), ''remote SDV'' (allowing SDV to be performed centrally), or hybrid approaches. 1,[4][5][6][7][9][10][11][12][13][14][15] The TransCelerate Biopharma Inc guidance on risk-based monitoring estimated that, of all generated queries, the percentage generated through SDV is as low as 7.8% overall and 2.4% in critical data. 20 These numbers may suggest not performing SDV at all.…”

“…1 Despite not being required by regulators, SDV is commonly performed on all CRF entries (100% SDV), which is time consuming, costly, and known to identify errors that have little impact on the trial outcomes. [2][3][4][5][6][7][8] Consequently, its use has been a major topic of debate in discussions on the efficiency of clinical trial monitoring, and various alternatives have been proposed. 1,[4][5][6][7][9][10][11][12][13][14][15] One alternative approach (which we term random SDV) is to limit SDV to a random sample of CRF entries: Inspection of a sample may be sufficient to assess the quality of all data provided by an investigational site.…”

“…[2][3][4][5][6][7][8] Consequently, its use has been a major topic of debate in discussions on the efficiency of clinical trial monitoring, and various alternatives have been proposed. 1,[4][5][6][7][9][10][11][12][13][14][15] One alternative approach (which we term random SDV) is to limit SDV to a random sample of CRF entries: Inspection of a sample may be sufficient to assess the quality of all data provided by an investigational site. 1,[4][5][6][7]11 This view is supported by the Guidelines for Good Clinical Practice, which state that ''statistically controlled sampling may be an acceptable method for selecting the data to be verified.''…”

“…1,[4][5][6][7][9][10][11][12][13][14][15] One alternative approach (which we term random SDV) is to limit SDV to a random sample of CRF entries: Inspection of a sample may be sufficient to assess the quality of all data provided by an investigational site. 1,[4][5][6][7]11 This view is supported by the Guidelines for Good Clinical Practice, which state that ''statistically controlled sampling may be an acceptable method for selecting the data to be verified.'' (16p26) In addition, by reducing the monitors' SDV workload, time becomes available for supporting site staff in increasing the quality.…”

A Data-Driven Approach to Risk-Based Source Data Verification

“…Specifically, the 4 hypothetical risk-based monitoring approaches proposed by Tantsyura et al 11 were simulated and analyzed. Since they had proposed well-thought-out and defined scenarios that were readily replicated in simulation algorithms and had proposed expected results, these scenarios were a good test of investigating whether real data could be used to simulate reduced SDV scenarios.…”