FDA Complete Response Letters Citing Inadequate External or Historical Controls

FDA complete response letters that cited inadequate external or historical controls

FDA has repeatedly issued complete response letters (CRLs) for applications where efficacy depended on a comparison with a non-concurrent control, whether that was a natural history cohort, a disease registry, a literature-derived benchmark, or pooled historical trial data. A CRL is not a final rejection, and each of the four main examples below was later approved after the sponsor resubmitted. The Drugs@FDA action packages show that the objection is rarely "external controls are unacceptable." It is almost always that the specific comparator was not shown to be comparable, was not prespecified, measured outcomes differently, or lacked the patient-level data FDA needed to check the analysis.

The clearest examples are narsoplimab (BLA 761152), palovarotene/Sohonos (NDA 215559), telavancin/Vibativ for nosocomial pneumonia (NDA 022407) and the Nexium IV rebleeding supplement (NDA 021689/S-014). All four CRLs were later resolved: FDA approved narsoplimab in December 2025, Sohonos in August 2023, Vibativ for HABP/VABP in June 2013 and the Nexium IV supplement in March 2014 ^789792793794. Two older actions from before the CRL format (Kytril and Synercid) and the Lumizyme post-CR registry analysis show the same reasoning.

Summary of key actions

Narsoplimab (BLA 761152): literature control, then a registry control

The 2021 CRL. FDA concluded the original BLA failed to establish substantial evidence of effectiveness. The application rested on a 28-patient single-arm study plus a literature-based historical control used to justify the assumed untreated response rate ⁵²⁵. FDA's review of that literature control identified:

• Unknown data validity and accuracy, endpoint measurement error, inconsistent response definitions and non-uniform definitions of 100-day survival across publications ³⁹⁵.
• Limited confounder adjustment, non-standardized treatment-visit timing, convenience sampling and likely selection bias ³⁹⁵.
• An assumed 15% untreated response rate supported only by the applicant's expert opinion ³⁹⁵.
• No characterization of calcineurin inhibitor withdrawal or modification, a common part of TA-TMA management ⁵²⁸.

FDA said the application lacked confirmatory evidence ¹⁰³. It described possible paths forward: a new randomized trial, or a single-arm trial with agreed responder criteria compared against a well-matched historical control and/or a prespecified threshold derived from an independent literature assessment. Alternatively, a resubmission could address the deficiencies with evidence of increased survival compared with an appropriate historical control ¹⁰³.

The registry-based resubmission. The sponsor then proposed the Japanese KSCTG transplant registry as an external control. FDA issued an Incomplete Response letter on April 28, 2025, rather than treating the resubmission as a complete response to the 2021 action ²⁸. FDA had advised that the KSCTG-based approach might not resolve the CRL deficiencies without individual patient data. It requested raw patient-level data, diagnostic algorithms and programming code to verify which registry patients met TA-TMA criteria. The applicant said the overall registry did not collect the data needed to verify or exclude a TA-TMA diagnosis ²⁸.

FDA's later review set out why the registry comparison was weak:

• Temporal mismatch. The registry covered 2000 to 2016 while treated patients enrolled from 2015 onward. This left a 3-to-7-year gap in which improvements in supportive care could favor narsoplimab ²³.
• Geographic and practice differences. The control was exclusively Japanese while the treated population was global ⁵²²³⁹³.
• Different disease timing and selection. Time from transplant to TA-TMA diagnosis was much earlier in registry controls ⁵⁸. Trial eligibility excluded early-onset, rapidly progressive and poor-prognosis patients, which produced a healthier treated population ³⁹³.
• Incomplete adjustment. Key prognostic factors were left out of the propensity-score model, leaving residual confounding ⁵²⁷³⁹³. Poor propensity-score overlap meant some treated patients had no comparable controls ⁶³³⁹³.
• Missing registry variables. The registry lacked high-risk TA-TMA factors and the components needed to determine response ³⁹³.
• Index-date imputation. FDA considered it unlikely that the imputed "latent treatment day" for registry patients was missing completely at random ⁵²⁷.

The approval. FDA approved narsoplimab-wuug (YARTEMLEA) on December 23, 2025 for adult and pediatric patients two years of age and older with TA-TMA ⁷⁸⁹. The March 26, 2025 resubmission had added a response to the CRL deficiencies, external control data and compassionate-use expanded-access data, and the review team recommended approval "based on the data included with this resubmission" ⁷⁹⁰. The approval review found substantial evidence of effectiveness in TMA-001 (17 of 28 patients responded), set against a historical response rate of 23.3% derived from 149 untreated patients in the published literature, with confirmatory evidence from expanded-access patients who had patient-level data ⁷⁹¹.

Sohonos, palovarotene (NDA 215559): natural history external control

FDA's Complete Response letter was dated December 23, 2022 ^593605606. It stated that the anticipated benefit of preventing new heterotopic ossification had not been established. It named two major unresolved issues: the "acceptability of the external control group (the Natural History Study, NHS) for evaluation of efficacy of the chronic/flare-up dosing regimen," and reliance on post-hoc analyses ²⁴.

Before the NDA, FDA had said the NHS might be acceptable as an external control but that data acceptability and comparability would be review issues. FDA also said randomized controls were preferred ⁴⁶⁷. The review identified four groups of problems:

• Comparability. The treated and NHS populations differed on multiple demographic and disease measures ⁴⁷⁰. The NHS-only group was older (19.9 vs. 14.1 years) and had higher baseline heterotopic ossification volume, more affected body regions and worse functional scores ²⁴⁷.
• Ascertainment. Flare-ups were reported at 0.15 vs. 0.07 per subject-month in Study 301 vs. the NHS ⁴⁶⁶. NHS contact was mainly every six months, compared with every three months (and every four weeks during flare-up dosing) in Study 301, and only Study 301 used daily symptom diaries ⁴⁶⁶. Among 39 crossover patients, 0.6 flare-ups were recorded prospectively in the final NHS year, but patients retrospectively reported 1.1 for the same period when they enrolled in Study 301 ⁴⁶⁵.
• Data collection. Worsening of flare-ups was not systematically captured in the NHS ⁴⁶⁵. CT scans were annual in the NHS vs. every six months in Study 301, and adverse events were not recorded in the NHS ⁴⁶⁸.
• Analysis. The external-control analyses were post hoc, used non-validated measures and had unequal follow-up ⁴⁷⁰⁴⁷¹. The prespecified primary analyses in Study 301 and the Phase 2 program had failed ⁴⁷⁴.

The CR page itself did not set out a detailed remedial study requirement beyond adequate support for efficacy ²⁴. Earlier FDA feedback had asked that every effort be made to use comparable assessment methods in the trial and the natural history control ⁴⁶⁷. Ipsen resubmitted on February 16, 2023 ⁴⁷⁵. After a June 28, 2023 advisory committee, where most members found effectiveness despite the acknowledged limitations of the NHS control ⁴⁷⁷, FDA approved Sohonos on August 16, 2023 ⁴⁶⁵⁷⁹².

Vibativ, telavancin (NDA 022407): historical controls in nosocomial pneumonia

FDA issued Complete Response letters for the nosocomial pneumonia/HABP indication on November 23, 2009 ⁶⁴¹ and December 21, 2010 ⁶⁵⁴⁶⁵⁹. A later February 2013 CR concerned manufacturing/CGMP, not the clinical deficiencies ⁶⁴⁹⁶⁵².

The sponsor compared telavancin-treated Phase 3 patients with historical patients who had received inadequate, inappropriate or delayed therapy. FDA concluded that "the baseline characteristics of the patients in the telavancin trials patients are not comparable to those in the historical control groups" ²². FDA said a comparability assessment would have to address age, APACHE-II score, percent ventilated, pathogens, adjunctive medications, ancillary care and management. It also said uncertainty about comparability justified discounting any noninferiority margin derived from the comparison ⁶⁴⁶⁶⁵⁶. Other limitations FDA cited were differences in study design, prognostic factors, pathogen prevalence, mortality reporting periods and changes in standard of care over time ⁶⁴⁶⁶⁵⁶. FDA rejected post-hoc selection of prognostic covariates as data-driven ²²⁶⁴⁶. It also found pooling of Studies 0015 and 0019 problematic because of differing mortality risk factors ²²⁵⁰⁴.

On historical evidence more broadly, FDA stated that published data supported noninferiority interpretation in NP/VAP only with all-cause mortality as the primary endpoint, not clinical response ⁶⁵¹. In those CR actions, FDA said it needed two adequate and well-controlled HABP trials ⁵⁰⁴. These were to use radiographic and clinical entry criteria, independent chest-radiograph interpretation, uniform specimen-quality criteria, minimized adjunctive antibacterials and standardized renal follow-up ⁵⁰⁴⁶⁴⁰.

FDA ultimately approved the HABP/VABP indication on June 21, 2013, limited to infections caused by susceptible Staphylococcus aureus when alternative treatments are not suitable ⁷⁹³. After a November 2012 advisory committee majority recommended that limited approval, FDA found on review of the July 2012 resubmission that the clinical and statistical issues with the two registrational HABP/VABP trials were substantially resolved ⁶⁵².

Nexium IV (NDA 021689/S-014): observational and literature comparisons

FDA issued a Complete Response on November 26, 2008 ⁵³⁹. The sponsor had supported a single adequate and well-controlled trial with observational studies and literature reviews. FDA said these were not adequate to serve as primary evidence of efficacy ⁵⁵⁸⁵⁴⁰. It cited lack of control and randomization, possible baseline differences and selection bias, ex-U.S. conduct and use of a comparator with unestablished efficacy ⁵⁵⁵. The historical Lau omeprazole trial was a single-center Hong Kong study with a placebo rebleeding rate of 20%, compared with 10% in the Nexium pivotal study, and FDA found no clear explanation for the difference ⁵⁵¹. FDA required an additional adequate and well-controlled trial ⁵⁴⁴. FDA approved the supplement on March 4, 2014 for risk reduction of rebleeding of gastric or duodenal ulcers following therapeutic endoscopy in adults. The approval letter records a further action letter on June 16, 2011 and a December 14, 2012 resubmission that responded to it ⁷⁹⁴.

Related actions showing the same reasoning

Lumizyme, alglucosidase alfa (BLA 125291). The February 27, 2009 CR cited CGMP, CMC, REMS and the lack of an agreed accelerated-approval verification study ^45638460. It did not reject a historical control. After the CR, however, FDA invited Pompe Registry data matching infantile-onset patients to the Myozyme historical-control cohort as a possible route to regular approval ⁴⁵⁶. FDA found that evidence limited for four reasons ^162453169:

• Only 25 of 48 patients were suitable for analysis.
• The comparator was non-concurrent.
• The registry mixed retrospective and prospective data, which could bias it toward survivors.
• There were no prospectively designed endpoints or statistical analysis plans.

Those limitations meant the data could not support approval for all ages ¹⁷⁹.

Pre-CRL era actions. Kytril (NDA 20-305/S-001) received a not-approvable letter on October 16, 1996. Comparisons with historical granisetron and prochlorperazine arms did not show that either regimen was effective. FDA asked for analyses against an appropriate historical control showing the regimens worked ⁷⁸⁸. For Synercid's VREF bacteremia indication (approvable action, March 5, 1998) ⁷⁶⁵, FDA found that published literature could not supply a reliable historical control. The reasons were differing regimens, infection definitions, endpoints and small sample sizes ⁷⁵⁹⁷¹⁰. FDA required a randomized, prospective confirmatory trial ⁷⁶³⁷⁶¹.

Recurring deficiencies FDA identifies

Across these actions, FDA's objections fall into a consistent set of categories:

1. Population non-comparability. Baseline imbalances in age, disease severity, prognostic factors or geography (narsoplimab ²³⁵²², Sohonos ²⁴⁷, Vibativ ²²).
2. Temporal drift in standard of care. Registry or literature periods that predate the treated cohort (narsoplimab ²³, Vibativ ⁶⁴⁶).
3. Unequal outcome ascertainment. Different visit schedules, diaries, imaging frequency or endpoint definitions (Sohonos ⁴⁶⁶⁴⁶⁸, narsoplimab literature control ³⁹⁵).
4. Post-hoc or data-driven analyses. Covariates or comparisons selected after unblinding (Sohonos ⁴⁷⁰, Vibativ ²²).
5. Inadequate statistical adjustment. Omitted prognostic covariates and poor propensity-score overlap (narsoplimab ⁵²⁷⁶³).
6. Missing patient-level data. Registry or published data that FDA cannot verify (narsoplimab ²⁸³⁹³, Lumizyme ¹⁶⁹).
7. Unvalidated thresholds. Assumed untreated response rates not grounded in independent evidence (narsoplimab ³⁹⁵).

FDA's statistical reviewers have framed a related concern using ICH E10. In the Nexium pediatric review (NDA 022101), FDA noted that a baseline-controlled study implicitly relies on an external control or efficacy threshold, so "the validity of the external control is crucial." It said the expected effect size should ideally be prespecified and justified ⁴⁴²⁴³⁸. That review was not tied to a CRL. The action on that application was an approvable letter driven by cardiac safety ⁴⁴³.

Practical implications for sponsors

The remedies FDA asked for across these actions are consistent. In rare diseases FDA remains open to non-randomized designs where a well-matched historical control or an independently derived, prespecified threshold is agreed in advance ¹⁰³. It also expects comparable assessment methods between the trial and the control cohort ⁴⁶⁷. It asks for individual patient data and code sufficient to verify eligibility and outcomes ²⁸. Where comparability cannot be shown, FDA has asked for a concurrent control: additional adequate and well-controlled trials ⁵⁰⁴⁵⁴⁴ or a randomized confirmatory trial ⁷⁶³.