Clinical Research
The data management plan and statistical analysis plan written before enrollment begins determine the quality and credibility of the data that comes out the other end. By the time a study closes, it is too late to fix a poorly designed database or an analysis plan that doesn't match FDA's expectations for your device type.
Build Your Data ProgramData Infrastructure
The case report form (CRF) is the instrument through which clinical observations are transformed into analyzable data. CRF design decisions made early in the study affect data completeness, query rates, and the effort required to produce the analysis-ready dataset at study end. A CRF that captures data in free-text fields where structured choices would suffice, or that collects data at the wrong time points relative to the protocol, creates data management problems that can't be solved retrospectively.
For electronic data capture (EDC) systems used in FDA-regulated investigations, 21 CFR Part 11 compliance requirements apply. The EDC system must maintain audit trails, use controlled access, and ensure data integrity through validation procedures. We work with EDC platforms (Medidata Rave, REDCap, Oracle Clinical, and others) and can assess 21 CFR Part 11 compliance documentation for the platform being used. Where paper CRFs are appropriate — typically small feasibility studies — we design paper instruments that minimize transcription errors and support efficient data entry.
The data management plan (DMP) specifies how data will be collected, entered, validated, queried, and locked. For FDA-regulated device studies, the DMP also documents the data validation procedures and edit checks that ensure data integrity — this documentation is part of the trial master file and may be inspected by FDA. A DMP written after enrollment has started is a retrospective reconstruction of decisions already made; a DMP written before first patient in ensures those decisions are made deliberately.
Source data verification (SDV) procedures, query management workflows, and database lock procedures are defined in the DMP. The query resolution rate and time-to-resolution are indicators of data quality that FDA biostatisticians pay attention to when reviewing clinical data. Sites with high query volumes relative to enrollment may trigger sponsor audit activities or data exclusion considerations during FDA review.
Statistical analysis of medical device trial data is governed by the pre-specified statistical analysis plan (SAP). FDA's expectations for device clinical statistics include: analysis of the primary endpoint using the pre-specified test, confidence interval reporting alongside p-values, handling of missing data using pre-specified methods (LOCF, multiple imputation, MMRM as appropriate), subgroup analyses that were pre-specified, and safety analyses covering all adverse events and device deficiencies. Sensitivity analyses may be appropriate to test the robustness of the primary finding.
The statistical output for FDA submission takes the form of tables, listings, and figures (TLFs) that present the analyzed data in a format FDA biostatisticians can review efficiently. TLF shells are defined in the SAP; the final output is produced from the locked database using the analysis programs specified in the SAP. We produce submission-ready TLF packages matched to the clinical section of the 510(k) or PMA.
Get Started
FDA biostatisticians review not just the analysis but the data management procedures that produced it. We build data programs designed to produce clean, auditable, submission-ready data from the first patient enrolled.