In the spring of 2019, a junior data scientist at Nexora Technologies named Marcus Webb noticed something he wasn't supposed to notice. Sitting in front of three monitors in the company's San Francisco research hub, he was running a routine audit of Phase II clinical trial results for Nexora's flagship product — a continuous glucose monitoring device marketed as the most accurate in its class. The numbers, he would later tell colleagues, didn't match.
"Not by a rounding error," Webb wrote in an internal message recovered in documents obtained by The Undisclosed. "By a factor. Someone changed the standard deviation on Trial Group C. Not once. Systematically, across six reporting periods."
Webb filed an internal report. He was reassigned within two weeks. By December of that year, he had left the company under a non-disclosure agreement whose terms, sources familiar with the document say, included a clause prohibiting him from speaking to any regulatory body without prior written consent from Nexora's legal team.
This is the story of how that NDA was not an isolated event — but part of a pattern that extended to at least four clinical trials, two regulatory submissions, and one IPO prospectus that valued Nexora Technologies at $4.2 billion.
The Company That Could Do No Wrong
Nexora was founded in 2014 by Dr. Caroline Ashford, a Stanford-trained biomedical engineer with a talent for fundraising matched only by her talent for generating press coverage. By 2017, the company had raised $380 million across three funding rounds, secured partnerships with two of the three largest pharmacy chains in the United States, and been featured on the covers of both Wired and Fast Company.
The product — initially called NexSense, later rebranded as ClearGlucose Pro — promised continuous glucose monitoring with clinical-grade accuracy at a consumer price point. For the estimated 37 million Americans living with diabetes, it represented something genuinely transformative: real-time blood sugar data without the finger-prick tests that defined daily life.
Nexora's clinical trials told a compelling story. In its pivotal Phase III trial — the results that would anchor its FDA submission — ClearGlucose Pro demonstrated a mean absolute relative difference (MARD) of 7.3%, placing it among the most accurate devices ever tested. Competing products averaged 9–11% MARD. The press releases wrote themselves.
What those press releases did not include: a parallel internal dataset, obtained by The Undisclosed from a source with direct access to Nexora's data infrastructure, showing a MARD closer to 11.4% when measured against the pre-registered trial protocol's unmodified data.
The Documents
Over fourteen months, The Undisclosed collected and analyzed 847 pages of internal documents, including raw trial data exports, version histories of statistical analysis plans, internal Slack logs, and email chains between Nexora's biostatistics team and its regulatory affairs department. We cross-referenced these against the company's public FDA submissions, its IPO S-1 filing, and four peer-reviewed papers published by Nexora-affiliated researchers.
The documents paint a consistent picture. In each case, the divergence between internal and externally-reported figures follows the same directional logic: accuracy metrics are improved, adverse event rates are reduced, and outlier patient data — the readings that most damaged the device's apparent performance — are reclassified or excluded from final analysis under justifications that were never registered in the original trial protocol.
"The exclusions weren't random. Every single exclusion improved the headline number. Every single one. In a legitimate dataset, you'd expect some to go the other way. That's how noise works."
— Former Senior Biostatistician, Nexora Technologies (identity withheld)
The version histories are particularly telling. Statistical analysis plans — the pre-registered documents that define how a trial's data will be analyzed — are supposed to be locked before data unblinding. They exist precisely to prevent researchers from adjusting their methodology after seeing results that disappoint them.
In three of the four trials for which we have documentation, Nexora's internal file systems show edits to the statistical analysis plan occurring after the trial's unblinding date. Two of those edits were made by accounts belonging to senior members of the regulatory affairs team — not the clinical research team responsible for the trial.
The Regulator That Looked Away
In October 2020, five months before Nexora's IPO and eight months before the FDA cleared ClearGlucose Pro for commercial sale, the agency received an anonymous submission through its MedWatch whistleblower portal. The submission, a copy of which was obtained by The Undisclosed, described "material discrepancies" between Nexora's Phase III trial data and an internal dataset the submitter claimed to have direct knowledge of.
The FDA's response, documented in agency records obtained through a Freedom of Information Act request, consisted of a single logged entry: "Submission reviewed. No action warranted at this time." No follow-up. No inspection. No request for additional documentation from Nexora.
We submitted detailed written questions to the FDA's press office. The agency responded with a two-sentence statement declining to comment on specific regulatory decisions. We submitted a second set of questions. We received no response.
We also reached out to three former FDA reviewers, two of whom agreed to speak on background. Both described an institutional culture in which whistleblower submissions from industry sources were treated with skepticism as a default — dismissed as competitive intelligence operations or disgruntled employee complaints unless accompanied by documentation substantial enough to initiate a formal investigation.
"The bar isn't 'is there something worth looking at,'" one former reviewer told us. "The bar is 'can we justify the resource expenditure of a formal audit.' Those are very different bars."
What the Device Actually Does
The clinical accuracy question is not academic. For diabetic patients using ClearGlucose Pro to make insulin dosing decisions — as Nexora's marketing explicitly encourages — a device that reads 4 percentage points less accurately than claimed can produce dangerous outcomes.
A glucose reading of 80 mg/dL with an 11% error margin places the true value anywhere between 71 and 89 mg/dL. At the lower end of that range, a patient administering insulin on the basis of that reading risks hypoglycemia — low blood sugar that, in its severe form, causes loss of consciousness, seizures, and in rare cases, death.
We identified 23 adverse event reports in the FDA's MAUDE database — the public registry of medical device complaints — referencing ClearGlucose Pro and involving accuracy-related incidents between the device's commercial launch in late 2021 and the time of our publication. Of those, six describe hypoglycemic episodes in which the patient or a family member attributed the incident directly to a device reading that diverged significantly from a confirmatory blood test.
The question of causation
We are careful here. Adverse event reports are self-reported and unverified. Correlation between a device reading and a medical incident does not establish that the device caused the incident. Medical devices — all medical devices — generate adverse event reports. Six reports over two years, for a device with an installed base of several hundred thousand units, is not in itself alarming.
What is alarming is that patients and their physicians are making treatment decisions based on accuracy claims that our documents suggest are materially overstated. Whether those overstated claims have directly harmed anyone is a question for investigators with subpoena power. We do not have subpoena power. We have documents.
Sources and Corroboration
Seven former Nexora employees contributed to this investigation, at varying levels of directness and documentation. Four provided written materials. Three spoke only on the condition of strict anonymity and asked us not to describe their roles in terms that might narrow the field of identification.
All seven, independently and without knowledge of one another's participation, described the same general pattern: clinical data that was understood internally to be weaker than publicly reported, a culture of pressure around regulatory deadlines that discouraged internal dissent, and departures — voluntary and otherwise — of employees who raised concerns through internal channels.
We presented our findings to Nexora's communications team six weeks before publication and offered the company an opportunity to respond in detail. The company provided a statement, reproduced in full below, denying any wrongdoing and characterizing our sources as "disgruntled former employees acting in bad faith." The statement did not address any specific document or data discrepancy we raised.
"Nexora Technologies stands fully behind the integrity of its clinical research and regulatory submissions. All trials were conducted in compliance with applicable regulations and Good Clinical Practice guidelines. Any suggestion of data manipulation is categorically false and defamatory."
— Nexora Technologies, Official Statement, January 2025
What Happens Now
We have shared our documentation with two members of the United States Senate Health, Education, Labor, and Pensions Committee, two investigative journalists at national outlets, and one academic researcher who specializes in clinical trial integrity. We are not the last word on this story. We are, we hope, the first.
Nexora's stock has risen 340% since its IPO. Dr. Ashford was named to Forbes' list of the world's most powerful women in business in both 2022 and 2023. ClearGlucose Pro has an average rating of 4.6 stars on the Apple App Store, where 74,000 users have reviewed it.
Marcus Webb, the data scientist who noticed the numbers didn't match, is no longer in the medical device industry. We know where he is. We have not published it. His NDA is still in force.
Methodology note: All documents cited in this investigation have been independently verified for authenticity by a third-party forensic document examiner retained by The Undisclosed. Statistical analysis was reviewed by two PhD-level biostatisticians who are not affiliated with Nexora or any of its competitors. Their assessments are on file.
Discussion
loading...