Five organizations in four notifications.

FDA’s Notifications on Data Integrity page says FDA notifies industry and the public when it finds data integrity concerns across multiple studies. The page, current as of July 14, 2026, lists four notifications, none from 2026 [1].

They name five contract research organizations in India whose bioavailability and bioequivalence (BA/BE) studies supported FDA applications: Semler Research, Bangalore (April 20, 2016); Panexcell Clinical Lab, Navi Mumbai, and Synchron Research Services, Ahmedabad (September 16, 2021); Synapse Labs, Pune (June 18, 2024); and Raptim Research, Navi Mumbai (March 28, 2025) [2][3][4][5].

Earlier, in July 2011, FDA told companies that bioanalytical studies conducted at Cetero Research from April 1, 2005, to June 15, 2010, “may need to be repeated or confirmed” [6]. FDA’s 2010 inspections at Cetero’s Houston, Texas, site and a third-party audit had identified misconduct “including document falsification and sample manipulation” [6].

What FDA said it found.

Form FDA 483 observations, the form states, “do not represent a final agency determination” [7].

Semler Research, 2016

FDA’s 2015 inspection of Semler’s bioanalytical facility found “significant instances of misconduct,” including “the substitution and manipulation of study subject samples” [2]. FDA investigators found a spreadsheet on Semler’s server indicating that one subject’s plasma samples were substituted for another’s [7].

Panexcell Clinical Lab and Synchron Research Services, 2021

Both sites had FDA inspections between November 18 and 22, 2019, which FDA said were not associated [3][8]. In one Panexcell study, subject data “appeared to separate into two distinct populations.” By the firm’s own analysis, subjects 1 to 12 had a Cmax geometric mean ratio of 71.26% and subjects 13 to 24 had 141.80%. Neither group met bioequivalence; all 24 together did, at a ratio of 100.52% [9]. At Synchron, FDA found “16 subject pairs with nearly identical concentration-time profiles,” which it called “indicative of subject sample substitution” [10]. FDA held each firm “responsible for the creation of false data” [9][10].

Synapse Labs, 2024

Analyzing Synapse data from several applications, FDA found “significant unexplained anomalies in the data that indicate the data were falsified” [4]. In one study, original data that met the acceptance criteria were “rejected and replaced with reanalyzed data” [11].

Raptim Research, 2025

After surveillance and for-cause inspections from 2019 to 2023, FDA concluded that Raptim “created and caused the submission of falsified in vitro data” [5]. In in vitro permeation test (IVPT) studies, FDA stated that Raptim appeared to have reanalyzed or reused replicates from one skin donor and recorded them as another donor’s [12].

EMA and WHO reviewed several of the same sites.

The European Medicines Agency (EMA) reviewed Semler, Panexcell, Synchron and Synapse under Article 31 of Directive 2001/83/EC. Each ended in a European Commission decision, after a CHMP recommendation to suspend medicines that relied on the site’s bioequivalence studies [13][14][15][16].

WHO issued notices of concern on Semler and Panexcell [17][18]. After a routine May 2024 inspection of Raptim covering clinical and analytical study data for three WHO applications, WHO considered those studies “to have been conducted at an acceptable level of compliance with WHO GCP/GLP/BE guidelines” [19].

FDA rows: year of the notification. EMA rows: year of the European Commission (EC) decision. WHO row: year of the notice of concern.

What the cases share.

Four problems recur, stated here in the agencies’ terms.

1. Profiles that should not match. Nearly identical concentration-time profiles across different subjects appear in the Semler, Panexcell, Synchron and Synapse letters [9][10][11][20]. At Raptim the same pattern appeared across skin donors [12].
2. Subgroups that fail apart and pass together. At Panexcell and Synchron, two subject groups each failed bioequivalence while the full study passed [9][10]. A Synapse study met the criteria “only when the data from group 4 are considered” with groups 1 to 3 [11].
3. Results rejected, replaced or excluded. At Synapse, replacing valid data with reanalyzed data changed the study result [11]. At Raptim, FDA cited the “retrospective rejection of acceptable study data and inappropriate use of the internal standard response,” including rejections based only on the area response of calibration curve standards (CCs) and QCs when the validated method relies on the analyte-to-IS response ratio. Its concern was that these practices “allow testing into compliance” [12]. Semler’s Form 483 states that a study report excluded 19 aqueous humor samples from PK analysis as blood-contaminated, while emails with the sponsor documented only 2 [7].
4. Records that cannot reconstruct the study. Semler’s Form 483 cites instrument audit trails not captured for archived projects from January to May 28, 2015, on three LC/MS instruments, analysts able to delete raw data folders, worksheets found “crumpled/torn in a trash pile” and calibrators and QCs that could not be reconciled [7]. FDA cited “uncontrolled and loose source documentation” at Synchron [10] and said Synapse had not provided “real-time documentation and study logs” [11].

What sponsors and applicants had to do.

Sponsors of approved and pending applications had to repeat the affected BA/BE studies at another site [2][3][4]. From 2021 that applied when the studies were “essential for approval,” and the new site could not have “known unresolved data integrity concerns” [3][4]. For Raptim, in vitro studies follow that rule. In vivo studies are handled case by case, with steps that may include justifying reliance on the data, reanalyzing the samples and redoing the statistics, or running a new study [5].

Companies were asked to respond within 30 days [4][5][8][21]. For Synapse, holders of affected approved or tentatively approved ANDAs, and for Raptim, holders of approved ANDAs and NDAs that relied on its in vitro studies, were asked for plans to re-conduct the studies and submit “within the next year,” or to request withdrawal [4][5]. Affected approved generics were rated BX [2][3][4][5]. A BX drug “is still approved and can be prescribed but is not recommended as automatically substitutable” [3][4][5]. In each case FDA stated that it had not identified safety concerns or signals for the affected products [2][4][5][8].

After the Cetero notification, an ANDA was rated BX in 2017 and its approval withdrawn in 2020 because the holder “repeatedly failed to submit the required data” [6][22]. In the EU, the CHMP initially found sufficient data for around 35 of the more than 400 medicines Synapse tested, and for one more on re-examination [16].

Record controls a CRO and a sponsor can show.

FDA’s April 2024 draft guidance on BA/BE data integrity places “the ultimate responsibility for the quality and integrity” of the data with applicants. As of October 26, 2026, FDA still lists it as a draft, not for implementation [23]. FDA’s 2018 question-and-answer guidance addresses CGMP records; read it with that scope in mind [24].

1. Attribute every action. Use unique log-in credentials “so that actions are attributable to a specific individual” [23]. With shared logins, “a unique individual cannot be identified,” and administrator rights belong with people independent of those responsible for the record content [24].
2. Keep audit trails on and read them. For closed systems, 21 CFR 11.10(e) requires secure, time-stamped audit trails, and “Record changes shall not obscure previously recorded information” [25]. Sites should restrict the ability to disable audit trails [23], and reviewers should read them with the rest of the record [24].
3. Keep the original and every result. If chromatography is reprocessed, “each result” is retained for review, and an invalidated result stays in the record with its investigation [24]. ICH M10 asks that original and reintegrated chromatograms be kept and reintegrations listed with reasons [26].
4. Set reanalysis and exclusion rules first. ICH M10 says reasons for reanalysis should be predefined, rules out reanalysis for a pharmacokinetic reason in comparative BA/BE studies, and asks for each reanalyzed sample’s initial value, reason and final accepted value [26]. The draft guidance recommends SOPs “with prespecified objective criteria” for repeat analysis and reintegration [23]. Invalidating a CGMP result requires “a valid, documented, scientifically sound justification” [24].
5. Lock methods and settings. Restrict changes to data or testing methods “by technical means where possible” [24]. At Raptim, FDA said earlier versions of the repeat analysis SOP “facilitated the inconsistent treatment of CCs and QCs,” and noted Raptim’s statement that it had since revised the procedure [12].
6. Sponsors: monitor the data flow. Plan monitoring around “the entire data-flow,” including steps where people intervene in what is recorded or retained, and audit sites against that plan [23].

Where WLTR fits.

WLTR verifies. Your lab validates. WLTR is a verification layer for the calibration data your CDS already produces. It does not replace your CDS, LIMS or method validation, and it does not detect manipulation. Your validated method and SOPs still decide which model is reported.

WLTR reads the raw instrument export directly and records a SHA-256 hash of the original file. It evaluates every candidate calibration model against the same acceptance criteria at once and keeps the full evaluation, so there is no curve to shop for and no failed attempt to hide [27]. Each exclusion is recorded with the user who made it and the reason given, and travels with the record into QA approval. Method configurations are frozen once results exist. Named users, role-based access and role-separated sign-off attribute every action, and review by exception gives QA a complete trail from export to approval.