Drug Safety and Pharmacovigilance Editing and Proofreading Services

During an inspection reported in a March 2026 warning letter, the Food and Drug Administration found a patient identifier in the source documents of a case a company had already invalidated.¹ The company's stated reason was that it "failed to capture the patient identifier."¹ The information existed in the papers the company held. It had not reached the record the company built from those papers.

Four elements decide whether a case can be submitted

FDA's E2D(R1) guidance states a minimum. "In order to be eligible for submission to any regulatory authority, ICSRs should have at least these minimum criteria."² They are at least one adverse event or adverse drug reaction or other observation, at least one suspect or interacting medicinal product, an identifiable patient, and at least one identifiable reporter.²

An older draft guidance is blunter about the consequence of a gap. If any one of the basic elements remains unknown after being actively sought, the draft says the report should not be submitted at all.³ Reports without such information "make interpretation of their significance difficult, at best, and impossible, in most instances."³ The same draft rules out wording too general to act on. A report saying that a patient "experienced unspecified injury" should not be included until something more specific can be determined.³

Actively sought is the operative phrase, and the March 2026 warning letter shows what happens when the seeking stops. A written procedure at that company "did not require follow-up on reported ADEs if consent was not obtained from the reporter."¹ The agency's answer was that "PADE regulations do not require the obtaining of consent to acquire additional information."¹

The narrative is held to written standards

FDA's E2D(R1) guidance, issued in March 2026, describes what a case narrative has to do. Its objective is "to summarize all relevant clinical and related information," and it "should serve as a comprehensive, stand-alone 'medical story.'"² Information "should be presented in a logical time sequence; ideally this should be presented in the chronology of the patient's experience, rather than in the chronology in which the information was received."² Abbreviations and acronyms "should be avoided, with the possible exception of laboratory parameters and units."² In follow-up reports, "new information should be clearly identified."²

A separate FDA guidance on good pharmacovigilance practices lists eight elements a good case report contains.⁴ They run from a description of the events and the time to onset of signs or symptoms to any other relevant information.⁴ Each of the eight has to be located in material a reporter supplied, then set down in an order nobody supplied.

What the coded fields lose has been measured

Researchers have applied natural language processing to FDA's adverse event reporting system. One study looked at age. Across reports from 2002 to 2018, the share missing that value in the structured field "increased from 21.9 to 43.8%."⁵ Reading the narratives recovered much of it. The tool "identified age for an additional one million ICSRs (10% of the total number of ICSRs from 2002 to 2018)" and cut the missing share to 27 percent overall.⁵

A second study, by scientists in FDA's Center for Drug Evaluation and Research, covered 15,321,967 reports filed between 1968 and 2020.⁶ Gender was missing from the structured field in 11.0 percent of them.⁶ Narrative text supplied a gender value for 472,000 reports whose coded field had none.⁶

That second study is also a check on any claim that a narrative preserves everything. Weight, ethnicity, and race were missing from the structured field in 77.9 percent, 99.4 percent, and 99.2 percent of reports.⁶ Among those, more than 95 percent were missing the same information from the narrative.⁶

Duplicate reports: matching prose, disagreeing fields

A 2025 study in the journal Drug Safety sampled 2,297 reports and classified 901 of them, or 39 percent, as duplicates falling into 237 unique duplicate sets.⁷ Within those sets, only 16 percent showed "concordance of all 7 data elements."⁷ The narratives behaved differently. The "median similarity score for the duplicate pairs was 0.87 compared to 0.48 for non-duplicate pairs."⁷ Inside a duplicate set the prose stayed close and the coded fields did not.

The sponsor's judgments, and the prose around them

Seriousness, expectedness, and causality are settled by the sponsor's safety physicians and regulatory reviewers. We neither make them nor check them. Everything else on the page is text.

In a case narrative we read for the order of events, for an abbreviation used before it is expanded, and for a date in the prose that disagrees with a date in the structured fields. We read for one spelling of a product name held across a submission. A periodic report is itself required to contain "a narrative summary and analysis of the information in the report," so the same reading applies at the aggregate scale.⁸ Client submissions are confidential throughout the job.

Cases arriving from other countries are written under a rule that shapes the work. FDA guidance asks that all report data elements be completed in English, and it names two exceptions.⁹ They are the reaction as reported by the primary source in the native language, and the case summary and reporter's comments in the native language.⁹ An English narrative is therefore built alongside the reporter's own words. The electronic format is bounded almost everywhere. The two comment fields that accompany the narrative carry ceilings of 20,000 characters each, and FDA's compatibility appendix recommends that systems "support the case narrative section with no size limitation."¹⁰