When Code Coverage Becomes an FDA Issue - Five Enforcement Cases Decoded
?Before ESL
- • FDA Warning Letter (Xoran, 2024): coverage not tied to software risk - three corrective responses rejected as inadequate
- • Medtronic PMA deficiency: branch coverage and configuration-specific testing questioned - FDA asked which branches, which config
- • Medtronic dead-code review: 100% reported coverage but FDA challenged dead code, ROM/RAM variants, and timing effects underneath
- • Biotronik PMA deficiency: firmware and hardware testing entangled - no software-specific regression test set
- • Medtronic MyCareLink: Major Deficiency Letter blocking approval - multiple software V&V and wireless verification gaps
- • Embedded targets (64–256 KB RAM) cannot fit full instrumentation - standard coverage approach does not work
VAfter ESL
- • ESL deployed Parasoft C/C++test module-by-module on-target instrumentation - overhead bounded by single module, not full firmware
- • Risk-tiered coverage: MC/DC for safety-critical modules, branch for moderate-risk, statement for low-risk - exactly what FDA expects
- • Coverage aggregation: unit, system, host, and manual test coverage merged into one unified project report
- • Dead code explicitly identified and classified per module - not hidden inside a 100% number
- • SensorDemo validation: 29 auto-generated unit tests, 83% line coverage, uncovered code traced to real defects (null-pointer, assignment bug)
- • Audit-ready evidence: per-module coverage + risk classification + uncovered-code justification + requirements traceability + configuration ID
Result: ESL’s approach transforms code coverage from a percentage chase into a defensible regulatory argument: what was tested, what was not, why the gaps are acceptable, and how evidence maps to software risk. Demonstrated on SensorDemo with 29 auto-generated tests finding real defects - the kind of evidence package that answers FDA questions before they are asked.
Bring ESL the software - not a cleaned-up demo
We will build, clean and connect the evidence. Start with a software evidence scoping workshop.
Schedule Scoping Workshop →