In Part 1 of this blog series, we explored how Statistical Data Analysis Centers (SDACs) support trial integrity through randomization oversight, event tracking, vendor collaboration, and regulatory support.

In this second part, we shift focus to the SDAC’s expanding role in safety surveillance, adaptive trial design, and ongoing risk assessment. These are critical areas to maintain sponsor blinding while enabling informed decision-making.

Firewalled communication models between SDACs and Safety Assessment Committees (SAC)

Many regulatory agencies want safety signal updates without being overwhelmed by hundreds of individual serious adverse events (SAEs) or suspected unexpected serious adverse reactions (SUSARs). They are primarily interested in events occurring at higher rates on an investigational product compared to control arms or historical rates. While sponsors can readily provide safety signal information for completed studies, incorporating active studies into the calculation can be more challenging.

An SDAC can identify signals that meet thresholds for further review or signals of particular interest, then alert a firewalled Safety Assessment Committee (SAC) within the sponsor. Data from ongoing and completed studies can be combined to generate a comprehensive set of outputs, ranging from SAE tables by treatment and by study to focused analyses of specific concerns such as cardiovascular events, interstitial lung disease, or Grade 4 neutropenia.

In some cases, the SAC would only be informed if specific statistical criteria were met. Such criteria could be based on blinded data (e.g., absolute rate of event meets a threshold, or lower limit of confidence interval of the absolute rate of events meets a threshold), or unblinded data (e.g., the relative risk of the event for patients on active compared to control meets a threshold, or the lower limit of the relative risk meets a threshold). Meeting a threshold would not necessarily require changes to study conduct but could prompt further investigation by the SAC (perhaps with ad hoc analyses performed by the SDAC) and possibly warrant notifying regulatory agencies or triggering an ad hoc DMC meeting.

A clear communication plan is essential. The SDAC-SAC approach allows the sponsor to stay blinded while enabling the SAC to access the results they need along with their knowledge of the treatment and regulatory discussions.

Managing adaptive trial designs while preserving study blinding

There are statistically sound approaches that allow for sample size adaptation, either to increase or decrease sample size for cost efficiency and/or to maintain statistical power. Some analyses can be conducted without knowing randomization assignments or the current treatment effect. Analyses can be done directly by the sponsor or the SDAC using their access to the necessary software.

More complex adaptations rely on knowledge of randomization and the current treatment effect. The SDAC is well-suited to support this, and the DMC may also be involved. Some studies have the DMC focus on safety and non-inferential efficacy, whereas a different group, named by the FDA guidance on DMCs as the Adaptation Committee, is formed to formally review interim efficacy analyses such as sample size adaptation. This Adaptation Committee could consist of either members external to the sponsor or a small, internal, firewalled group that reviews outputs created by the SDAC (e.g., conditional power) to make decisions on sample size adaptation.

For such designs, it is important that adaptation decisions be communicated in a way to minimize the information that might be inferred. Therefore, careful communication planning with respect to the information conveyed after an interim analysis is crucial. Alternative approaches to prevent the backward calculation or reverse engineering of treatment effect estimates can be achieved by carefully designing the adaptation rules. For example, if the sample size is increased after interim analysis, trial sites could be informed that the targeted enrollment number has not been reached yet rather than the exact new sample size. Alternatively, one could use step functions for sample size increase rather than a continuous sample size increase rule to limit inferred knowledge on the target sample size. Knowing that the study will go beyond its originally intended number of patients/events is informative, but hopefully minimally so. Even so, best practice is to keep the exact statistical methodology for the sample size reassessment on a need-to-know basis.

Periodic safety updates to DMC

As noted earlier, a full set of materials is provided to the DMC as part of a scheduled DMC meeting. Many DMCs also receive periodic safety updates, e.g., the narratives of important events, as their remit, including PDFs of SUSARs or narratives from CIOMS reports or SAEs. The intention is to alert the DMC to critical events that could trigger an ad hoc DMC meeting or additional information. Providing these periodic updates to the DMC is especially important if the time between DMC meetings is lengthy or if information is limited on how the new treatment impacts the studied patient population. The process for these periodic safety events should be specified in the DMC Charter.

These outputs might go to the full DMC, the DMC Chair, or the DMC members who have the clinical expertise to review them (e.g., not the DMC statistician). There might be formal written acknowledgment from the DMC Chair whether action was needed, although more frequently only when an action is taken. The DMC Charter might encourage that the frequency or filter change as the study database matures. For example, the frequency might start as monthly but then go to bi-monthly after the first year. Or, it might start by providing all SAEs, but after the first year only pass along fatal events.

These periodic safety event updates are typically provided to the DMC without treatment information. If requested, the SDAC can provide the treatment information, so it is important that the SDAC have access to randomization information in real time without needing to request it. If a subject was recently randomized and the DMC felt it was critical to obtain the treatment, the SDAC should be able to get it without alerting the sponsor. This would be important if the safety event was critical enough to warrant a DMC recommendation for action. The DMC can call an ad hoc meeting to review the case and propose action. The SDAC could manually unblind each incremental event or wait for the DMC request to unblind.

It is understandable that the DMC and the sponsor would want to keep the DMC apprised of individual important safety events that occur between DMC meetings. However, the DMC is at its most valuable when looking at by-arm results using cumulative data. There are typically very few types of events where one or two instances of the event would cause the DMC to have enough evidence to recommend action. Examples of a single event causing consternation might be anaphylaxis, PML, or Hy’s Law case — or a death in a reasonably healthy population.

Less common, but likely more helpful, is to pass along a tabulation of safety data periodically. For example, a table of SAEs presented by arm (using the proper denominators of all subjects at risk as of this time) accompanied by a listing of the incremental SAEs since the previous transfer could be helpful to the DMC. That would require having datasets provided to the SDAC, as opposed to PDF, that have the necessary information to compute an accurate denominator and treatment information for all participants at risk, not just those having at least one SAE.

The concept of participants at risk becomes increasingly complicated when data snapshots and different sources of data (i.e., randomization, clinical, and safety) are considered. There typically is no formal monitoring plan for these events, but some studies have implemented plans. For example, the DMC might start reviewing serious infections starting after the fourth serious infection is seen with specific guidance for when an ad hoc meeting should be called based on by-arm difference with statistical underpinnings to the criteria (e.g. ≥4 vs. 0, ≥6 vs. 1, ≥8 vs. 2). The SDAC would assist in receiving information in real-time, unblinding cases and working with the DMC. Again, this could be done at the time each new event is observed, or perhaps monthly.

Another approach would be to create profiles of patients that have had clinically important events. Again, this would require the SDAC to get a fresh transfer of data (likely including lab data as well as CRF data). These patient profiles would likely depict demographics, disposition, treatment exposure, AEs, and key lab results. Lab data might also be shown in a spaghetti plot (e.g., each subject in their own plot showing ALT and AST represented by “xULN” multipliers over time).

Final takeaways

As clinical trials become more complex, the role of the Statistical Data Analysis Center (SDAC) continues to expand beyond its traditional support for DMC meetings. Across both operational and strategic activities, the SDAC helps sponsors, DMCs, and regulators maintain trial integrity while preserving appropriate blinding.

In Part 1 of this two-part blog series, we explored how SDACs support randomization oversight, event tracking, vendor collaboration, and regulatory interactions. In Part 2, we examined their growing role in safety surveillance, adaptive trial design, and cumulative risk assessment.

Across all these activities, one theme remains consistent: careful communication planning is essential. Whether supporting interim adaptations, safety signal detection, or regulatory requests, the SDAC helps balance the need for actionable insights with the need to minimize operational bias.

As sponsors increasingly adopt adaptive designs and more sophisticated safety monitoring approaches, the value of an experienced and independent SDAC becomes even more important, both as an analytical partner and as a trusted steward of sensitive clinical trial data.

Interested in learning more?

Download our white paper, “DMCs for Oncology Studies”:

Share Via