AcceleraQA Regulatory Intelligence — Case Analysis
483-to-Warning Letter Crosswalk
with Integrity Impact
A finding-by-finding comparison of the Form FDA 483 issued to clinical investigator Mark S. Dacey, M.D. and the warning letter that followed fifteen months later — with integrity impact and escalation rationale for each observation.
Background
This document compares a Form FDA 483 issued to a clinical investigator with the FDA warning letter that followed about fifteen months later. Both source documents are public. FDA released the 483 through the Freedom of Information Act, and the warning letter is posted on FDA's website.
The Inspection
FDA conducted this inspection under its Bioresearch Monitoring program (BIMO), which is how the agency inspects the conduct of clinical research. This was a clinical investigator inspection — FDA examined how the investigator and site carried out the studies, protected subjects, and recorded data. The inspection ran October 23 to November 8, 2024, in Denver, Colorado (FEI 3031610524), covering two ophthalmology protocols. Redacted findings point to intravitreal injection studies of a retinal condition, based on references to injections, study-eye and fellow-eye designations, and intraocular pressure monitoring.
The Form FDA 483
A Form FDA 483 lists the observations an FDA investigator records at the close of an inspection. It documents conditions the investigator considers objectionable, and it is explicitly not a final agency determination of compliance. This 483 ran nine pages and raised three observations across the two protocols: Observation 1 concerned informed consent not obtained in accordance with 21 CFR part 50; Observation 2 concerned an investigation not conducted per the investigational plan under 21 CFR 312.60; Observation 3 concerned failure to prepare and maintain adequate and accurate case histories under 21 CFR 312.62(b). Together the three observations carried roughly two dozen sub-findings, ranging from masking and randomization problems to a broad set of source-to-EDC discrepancies.
The Investigator's Written Response
The investigator responded on December 3, 2024. The response was substantive: the site reported the matter to the medical monitor and the IRB, notified the affected subject, arranged destruction of the sample with sponsor confirmation, and proposed process changes including source-binder labeling and a consent-tracking feature in its clinical trial management system.
The Warning Letter
A warning letter is a formal notice that FDA has determined significant violations occurred. FDA's Center for Drug Evaluation and Research, through its Office of Scientific Investigations, issued this warning letter on February 2, 2026 — roughly fifteen months after the inspection closed. The letter cited a single violation: failure to obtain informed consent, under 21 CFR 312.60 and 21 CFR 50.20. It described an optional sample collection performed on a subject who had twice declined consent in writing, and stated that the investigator's written response did not adequately address how the investigator would ensure oversight of study procedures going forward.
How to Read This Document
Each finding is shown in three parts: the verbatim 483 text (shaded header bar), then two analysis columns. The Integrity Impact column asks what the finding does to the trustworthiness of the study and its data. The Escalation Rationale column asks why FDA did or did not carry the finding into the warning letter. Redactions appear as [redacted].
The Reversibility Test
The single most useful cut is between damage you can rebuild and damage you cannot.
Bias you can never remove once it has entered the data — unmasking of assessments and corrupted stratification cannot be repaired, only quarantined.
A missing lab you can retrieve, a discrepant value you can re-verify against source, a consent you can obtain again. Costs the sponsor time and monitoring.
The data may be retrievable while a structural consequence (sequence, eligibility, attribution) is already locked in.
Crosswalk: 483 Text, Integrity Impact, and Escalation
The shaded bar under each observation carries the verbatim 483 wording. Impact cells are tinted by reversibility. The cited finding is highlighted throughout.
Observation 1: Informed Consent on a Superseded ICF Version
| Finding / Frame / Disposition | Integrity Impact | Escalation Rationale |
|---|---|---|
|
Obs 1.a / 1.b — Superseded ICF version Participants were never consented with the most recent IRB approved version of the ICF. (a) Subject [redacted] was screened and consented on [redacted] with ICF Version date 12Oct2021 (IRB approved 10/14/2021); ICF Version date 17Aug2022 (IRB approved 8/18/2022) should have been presented. (b) Subject [redacted] was screened and consented on [redacted] with the same 12Oct2021 ICF version when the 17Aug2022 version should have been presented. 21 CFR 50.20Not cited |
Mixed Consent validity for two subjects. Remediable through re-consent and version control, but authorization for those subjects' data is in question until corrected. |
Both subjects did consent, just on a superseded ICF. FDA reserved its consent citation for the graver refusal-then-procedure pattern. |
Observation 2: Investigation Not Conducted Per the Investigational Plan — Protocol I
| Finding / Frame / Disposition | Integrity Impact | Escalation Rationale |
|---|---|---|
|
Obs 2.I.1 — Randomized before eligibility confirmed Subjects are randomized prior to the determination of eligibility by the Investigator. On [redacted], Subject [redacted] was randomized and IP kit # [redacted] administered. All screening laboratory results were not reviewed by the Investigator until 1/30/2024. 21 CFR 312.60Not cited |
Mixed Population & Contemporaneous. Labs are retrievable, but the sequence — dosing before eligibility review, labs unreviewed until 1/30/2024 — cannot be undone. Eligibility was unverified at the moment of exposure. |
Correctable at the data level and sponsor-monitorable. No standalone subject-harm nexus a CAPA could not close. |
|
Obs 2.I.2 — Wrong stratification arm The baseline [redacted] score for the study eye (OS) for Subject [redacted] was [redacted]. However, at randomization the subject was stratified to the group designated as [redacted] letters. 21 CFR 312.60Not cited |
Irreversible Efficacy set. Wrong stratification arm from an incorrect baseline study-eye score corrupts the randomization set and biases any stratified analysis. The imbalance cannot be removed after the fact. |
A data-integrity and analysis issue, not a rights violation. Absorbed into the inspection narrative. |
|
Obs 2.I.3 — Unmasked investigator performed assessments Per Protocol Section 6.3.2, Masking, the treatment administrator (Unmasked Investigator) is not permitted to perform pre-dose assessments (after baseline) and unscheduled safety assessment visits. Study records demonstrate all study assessments, including administration of study treatment for Subject [redacted] on [redacted], were completed by [redacted], who was designated as an unmasked Investigator on this study. 21 CFR 312.60Not cited |
Irreversible Efficacy set, highest severity. The unmasked investigator performing pre-dose and post-baseline assessments collapses the assessment firewall. Assessment bias is uncontrollable and unremovable once introduced. |
Design-integrity concern owned partly by site staffing and sponsor oversight, not the PI's non-delegable consent duty. |
|
Obs 2.I.4 — Late AESI reporting Per Protocol Section 8.3.8.1, sight-threatening adverse events are required to be reported to the sponsor no more than [redacted] after the Investigator becomes aware of the event. At the Early Term Visit ([redacted]), the [redacted] score for Subject [redacted] decreased greater than [redacted] letters in the study eye, but the AESI was not immediately reported to the sponsor. 21 CFR 312.64Not cited |
Reconstructable Safety database, Complete & Contemporaneous. A late-reported sight-threatening AESI is retrievable, but the reporting delay degrades the timeliness the safety signal depends on. |
Reporting-timeline issues resolve through process fixes and sponsor pharmacovigilance. |
|
Obs 2.I.5 — PRN decisions by coordinators Study records fail to sufficiently demonstrate the evaluation and decision for PRN treatment (Week 20 through Week 48) is completed by the masked Investigator / Sub-Investigator. Subject records document the evaluation of specified criteria were completed by individuals delegated as study coordinators. For example, Subject [redacted] (Week 24) was determined to have met criteria and received PRN treatment. 21 CFR 312.60Not cited |
Irreversible Efficacy & exposure-response. PRN retreatment decisions (Wk 20–48) made by coordinators rather than the masked investigator make the exposure record itself unreliable, which no reconciliation restores. |
A delegation-of-authority concern, correctable and monitorable in principle. |
|
Obs 2.I.6 — Unapproved local lab for eligibility testing Protocol required safety laboratory assessments were to be performed by the central laboratory. Study records document [redacted] testing for Subject [redacted] was instead performed by a local laboratory. There is no evidence this deviation was approved by the sponsor prior to determining the subject as eligible for enrollment. Protocol planNot cited |
Reconstructable Population & eligibility. Local-lab infection testing without an approved deviation can be re-verified against the local result, but eligibility rested on an unapproved method. |
Sponsor-monitorable protocol deviation. |
|
Obs 2.I.7 — Sample collected after documented refusal ⚠ CITED Subject [redacted] declined their consent to provide [redacted] samples for pharmacokinetic and pharmacodynamic assessment. However, an [redacted] sample was collected on [redacted] (Week 20). 21 CFR 312.60 → 50.20Cited in WL |
Irreversible PK/PD datum, consent-tainted. A sample drawn against a twice-documented refusal is ethically and analytically unusable. Both an unusable datum and a rights violation in one act. |
The survivor. FDA re-anchored it under 50.20 and 312.60 because the subject twice declined in writing and the invasive procedure was performed anyway, with ocular-risk exposure. |
|
Obs 2.I.8 — Post-dose IOP measurements outside required window IOP measurements (only in the study eye) were required to be taken [redacted] after administration of study treatment. These measurements were not always taken within this time frame, including: Subject [redacted] (Week 24); Subject [redacted] (Week 4); Subject [redacted] (Week 4, Week 12); Subject [redacted] (Week 20). Protocol planNot cited |
Reconstructable Safety set, Contemporaneous. Post-dose IOP checks outside the required window lose the time relationship to dosing, though the readings themselves exist. |
Timeframe-adherence issue, correctable and monitorable. |
Observation 2: Protocol II
| Finding / Frame / Disposition | Integrity Impact | Escalation Rationale |
|---|---|---|
|
Obs 2.II.1 — Ineligible subject dosed Subjects were excluded from enrollment with a known history of alcohol and/or drug abuse within [redacted] prior to Visit 1. Subject [redacted] was randomized on [redacted]. Medical records document a history of alcohol and methamphetamine use (September 2022). On the Eligibility Criteria worksheet, exclusion criteria (#8) was left unmarked; however the subject was still determined eligible and kit dispensed. Protocol planNot cited |
Mixed Population, leaning irreversible. An ineligible subject (unmarked exclusion #8 despite documented alcohol and methamphetamine use) was dosed. The subject should not sit in the analysis set; forces per-protocol exclusion. |
Serious eligibility failure, but resolves through source review and adjudication. |
|
Obs 2.II.2 — Randomization before signed eligibility Subjects are randomized prior to the determination of eligibility by the Investigator. On [redacted], Subject [redacted] was randomized into the study and kit assigned via IWRS. However, source records documenting the Investigator's review and confirmation of eligibility were not signed until 9/21/2022. 21 CFR 312.60Not cited |
Mixed Population & Contemporaneous. Randomization and kit assignment preceded signed eligibility confirmation. The sequence is fixed in the record. |
Sequencing and documentation deviation. |
|
Obs 2.II.3 — Missing Visit 1 blood draws According to Protocol Section 7.1.1, a blood draw was required for testing of both [redacted] and [redacted]; documentation of prior test results was not available. (a) No laboratory results available to demonstrate prior testing for Subject [redacted], no blood collected. (b) No laboratory results available for Subject [redacted], no blood collected. Protocol planNot cited |
Reconstructable Population & baseline, Complete. Missing Visit 1 blood draws with no retrievable prior results is a true data gap, not a transcription fix. |
Data-completeness gap, correctable only if source can be located. |
Observation 3: Failure to Maintain Adequate and Accurate Case Histories
| Finding / Frame / Disposition | Integrity Impact | Escalation Rationale |
|---|---|---|
|
Obs 3.I.1 — Spanish-speaking subject, undocumented consent process Documentation regarding the consenting process of Subject [redacted] (Spanish speaking only) on [redacted] is not sufficient to support the review of the English ICF versions for optional sampling and participation by the translator. No copy of the English versions of these consents were maintained in the study records; the note to file describing the consenting process was dated 1/7/2024. Translated versions of both optional consents were approved by the IRB at the time of consent but were not provided. 21 CFR 312.62(b)Not cited |
Mixed Consent authorization. Undocumented Spanish-language consent process with no English ICF retained puts the authorization basis in question; partly reconstructable through file reconstruction. |
Case-history completeness issue, remediable through records management. |
|
Obs 3.I.2 — Screening finding omitted from medical history [redacted] observed at the screening ([redacted]) for both eyes (OU) has not been reported as medical history (to include ocular) for Subject [redacted]. 21 CFR 312.62(b)Not cited |
Reconstructable Baseline & safety, Accurate/Complete. A screening finding (OU) omitted from medical history is a capture gap correctable against source. |
Data-capture discrepancy. |
|
Obs 3.I.3 – 3.I.5 — Source-to-EDC discrepancies across assessments Discrepancies noted between source records and EDC data for indirect ophthalmoscopy (optic nerve, retina, vitreous), slit lamp assessments (iris, lens, keratic precipitates), and refraction assessments — across multiple subjects, visits, and eyes. (The 483 details each in tabular form; summarized here.) 21 CFR 312.62(b)Not cited |
Reconstructable but systemic Whole-site reliability, Accurate/Consistent. Individually correctable, but the pattern across subjects and both protocols reads as systemic, inviting doubt over the entire site dataset. |
Classic ALCOA+ transcription gaps, central to the inspection story but not enforcement-defining. |
|
Obs 3.I.6 / 3.I.7 — Unrecorded concurrent ocular treatments Per Protocol Section 8, any ocular and/or systemic treatments during a participant's medical history before Day 1 should be recorded within the eCRF; a treatment (OD) received by Subject [redacted] was not reported. Per Protocol Section 8.1.3, any ocular procedures performed during the study will be recorded on the Concurrent Ocular Procedures Log; injections for two subjects (OD and OS) observed in medical records were not reported. 21 CFR 312.62(b)Not cited |
Mixed Safety attribution. Unrecorded prior and concurrent ocular treatments make confounders invisible; you cannot separate an IP effect from a co-intervention. Records exist, so data is reconstructable, but attribution is already muddied. |
Documentation-completeness gaps under Protocol 8 and 8.1.3. |
|
Obs 3.I.8 — Subject-reported AEs not entered in EDC Study records do not demonstrate all adverse events reported by subjects have been adequately reviewed and evaluated to assure accurate reporting within the EDC: 1) Subject [redacted]: eye twitching (OS) reported on 3/5/2024 (Week 20); 2) Subject [redacted]: mild burning in eyes after last injection reported (Week 16). 21 CFR 312.62(b)Not cited |
Reconstructable Safety set, Complete. Subject-reported AEs absent from the EDC undercount the safety database; recoverable if source captured them. |
Safety-data-capture gap, correctable. |
|
Obs 3.I.9 / 3.I.10 — Multiple IOP readings, no selection rule Multiple IOP measurements are taken without justification; there is no assurance the value reported within the EDC is the initial value obtained. For example, at Week 12 and Week 24 for Subject [redacted], multiple IOP measurements (OD) were recorded while a different number were recorded for OS. For the Week 12 visit for Subject [redacted], the EDC documents a reading obtained at 12:16 using one device, but the clinic note states two readings were obtained and the value reported in the EDC was obtained using [redacted] at 12:26. 21 CFR 312.62(b)Not cited |
Irreversible Data originality, Original/Attributable/Accurate. Multiple IOP readings without a documented selection rule mean no assurance the EDC value is the initial reading; the device and timestamp conflict (12:16 vs two readings at 12:26) cannot be reconciled to source. Originality fails at capture. |
Core data-originality concern, remediable prospectively through SOPs but not for the affected records. |
|
Obs 3.II.1 — Medical history date entered inaccurately Study related records for Subject [redacted] document positive testing for [redacted] in August 2014. However, this information is reported inaccurately within the EDC as starting in 2016 (Unknown). 21 CFR 312.62(b)Not cited |
Reconstructable Medical-history accuracy. A positive test documented August 2014 but entered as starting 2016 is a correctable accuracy discrepancy. |
Accuracy discrepancy. |
|
Obs 3.II.2 — Unscheduled-visit IOP readings not entered in EDC An unscheduled visit for Subject [redacted] was completed on [redacted]. During this visit, an IOP measurement was taken at 13:37 for both eyes. These results ([redacted] OD; [redacted] OS) have not been reported within the EDC. 21 CFR 312.62(b)Not cited |
Reconstructable Safety set, Complete. Unscheduled-visit IOP readings (both eyes) never entered in the EDC are recoverable from source. |
Completeness gap. |
Integrity Impact by Theme
Grouping findings by what they threaten shows where the irreversible damage concentrates.
| Impact Theme | Findings | Reversibility | Dataset at Risk |
|---|---|---|---|
| Efficacy-endpoint bias | 2.I.2 stratification, 2.I.3 masking, 2.I.5 PRN decider | Irreversible | Efficacy set |
| Population and eligibility | 2.I.1, 2.I.6, 2.II.1, 2.II.2, 2.II.3 | Mixed | ITT and per-protocol |
| Safety database | 2.I.4, 2.I.8, 3.I.6, 3.I.7, 3.I.8, 3.II.2 | Mostly reconstructable | Safety set |
| Data originality | 3.I.9, 3.I.10 | Irreversible | IOP source reliability |
| Systemic accuracy | 3.I.3, 3.I.4, 3.I.5, 3.II.1 | Reconstructable but systemic | Whole-site dataset |
| Consent-tainted | Obs 1, 2.I.7, 3.I.1 | Irreversible or re-consent | Authorization to use data |
ALCOA+ Scorecard
The site fails on nearly every non-legibility dimension. Representative findings shown; the list is not exhaustive.
| ALCOA+ Dimension | Status | Representative Findings |
|---|---|---|
| Attributable | Fails | Masking violation (2.I.3), wrong PRN decider (2.I.5), device and timestamp conflict (3.I.10) |
| Legible | Not implicated | No legibility findings in the 483 |
| Contemporaneous | Fails | Eligibility signed after randomization on 9/21/2022 (2.II.2), safety labs unreviewed until 1/30/2024 (2.I.1), after-the-fact notes to file |
| Original | Fails | Undocumented initial IOP value, no selection rule (3.I.9, 3.I.10) |
| Accurate | Fails | Source-to-EDC discrepancies (3.I.3 to 3.I.5), mis-entered test date (3.II.1) |
| Complete | Fails | Unreported AEs (3.I.8), unrecorded procedures (3.I.6, 3.I.7), missing labs and IOP (2.II.3, 3.II.2) |
| Consistent | Fails | Systemic source-to-EDC mismatch across both protocols |
| Enduring / Available | Not implicated | No retention or access findings in the 483 |
The Three Drivers Behind the Single Citation
The cited finding started as Observation 2, Protocol I, item 7 — a protocol deviation about a PK/PD sample. FDA reframed it as a Part 50 consent violation. Three factors drove that choice.
1. Documented Refusal, Then the Procedure Anyway
The subject declined consent for optional aqueous humor sampling on two dates in 2023, in writing. The site collected the sample from the study eye at the Week 20 visit in March 2024. A missing signature is a paperwork failure. Performing an invasive procedure on a subject who affirmatively declined, twice, is a rights violation. That distinction separates a 483 item from a warning-letter citation.
2. A Demonstrable Subject-Harm Exposure
Unlike the version-control and discrepancy findings, this one carried a direct safety nexus. FDA enumerated the risks the subject was needlessly exposed to: blurred vision, bleeding on the surface of or inside the eye, damage to the lens, and inflammation or infection inside the eye. A subject-welfare consequence converts a deviation into an enforcement priority.
3. A Response That Fixed the Mechanism, Not the Oversight
The investigator responded, and the CAPA was not trivial: self-report to the medical monitor and IRB, subject notification, sample destruction with sponsor confirmation, and proposed source-binder labels and a CTMS consent-tracking feature. FDA acknowledged all of it and still escalated — because the response did not explain how the investigator would ensure adequate oversight of study procedures, and gave no implementation or training detail. The trigger was not the absence of a CAPA. It was a CAPA that addressed the tool while leaving the investigator-accountability question under 312.60 unanswered.
Study-Level Bottom Line
Reconstructable findings can be remediated through source verification, re-consent, and query resolution — they mostly cost the sponsor time. The irreversible findings — unmasking of efficacy assessments and corrupted stratification — cannot be repaired, only quarantined.
If this site is a material contributor, realistic consequences run from per-subject exclusions and per-protocol trimming, to sponsor sensitivity analyses excluding the site, to FDA questioning the site's data reliability in the marketing application. The breadth across both protocols is the aggravating signal: it reframes the problem from several bad data points to a site whose quality system did not reliably produce trustworthy data.
Takeaways for BIMO and QA Training
- Reversibility, not severity, sets the true cost. The masking and stratification failures are the costliest here because bias, once in the data, cannot be removed.
- Escalation and integrity impact are different axes. FDA cited a consent violation, but the largest integrity damage sits in uncited efficacy-endpoint findings. A site can pass enforcement triage and still hand the sponsor unusable data.
- Systemic beats isolated. Scattered discrepancies get queried; a pattern across subjects and protocols invites doubt over the whole dataset.
- Response adequacy is itself a finding. A CAPA that repairs the mechanism but omits oversight, implementation, and training can keep a finding alive to citation.
Sources
- United States, Department of Health and Human Services, Food and Drug Administration. Form FDA 483, Inspectional Observations: Mark S. Dacey, M.D. FEI 3031610524, issued 8 Nov. 2024. FDA FOIA release 2026-7541.
- United States, Department of Health and Human Services, Food and Drug Administration. "Mark S. Dacey, M.D. - 722942 - 02/02/2026." FDA, 2 Feb. 2026, fda.gov/…/mark-s-dacey-md-722942-02022026.
- Prepared by AcceleraQA for training and QA use. Verbatim 483 text is transcribed from the FOIA release with redactions preserved as [redacted]; the tabular discrepancy findings (3.I.3 to 3.I.5) are summarized. Endpoint and design inferences follow from the 483's ophthalmology context and are not stated verbatim in the source.