Executive Summary

The Standard for the Exchange of Nonclinical Data (SEND) is the CDISC-endorsed standard for organizing nonclinical (toxicology, pharmacology, etc.) study data in regulatory submissions to the FDA. First released in 2011 (SENDIG v3.0) under SDTM v1.2, and subsequently updated (e.g. v3.1 in 2016 under SDTM v1.5, and v3.1.1 in 2021), SEND is explicitly built on the SDTM framework (^[1]) (^[2]). In practice, SEND provides an “SDTM-like” Implementation Guide tailored for animal studies, prescribing domains and variables that mirror clinical SDTM datasets but accommodate nonclinical study design (e.g. animals instead of human subjects, dosing periods, tissue findings, etc.) (^[1]) (^[3]). The FDA now requires all in-scope nonclinical studies (for INDs, NDAs, ANDAs, BLAs) initiated after specific cutoff dates to be submitted in SEND format or risk a refuse-to-file (^[4]) (^[5]). Key components of a SEND submission include dataset files (datasets like TS, TA, TX, DM, DS, SE, EX, and any applicable findings datasets), a metadata Define-XML file, and a nonclinical Study Data Reviewer’s Guide (nSDRG) (^[6]) (^[7]). Industry panels and FDA experts emphasize that the uniform format improves review efficiency and enables cross-study analysis of toxicology data (^[8]) (^[9]). For example, aggregated SEND data from thousands of animal studies has been used to quantify background lesion incidence in control animals, with 94% of microscopic-finding records in SEND v3.1 using standardized terms (^[10]) (^[8]). This report reviews the historical development of SEND, current FDA rules and timelines, the technical structure of SEND datasets, tools and processes for implementation, case examples of usage, and future directions for SEND and standardized nonclinical data submission.

Introduction and Background

Standardized Study Data has been required for clinical trials (CDISC SDTM and ADaM) for years. Under the Food and Drug Administration Safety and Innovation Act (FDASIA, 2012) and subsequent guidance, the FDA mandated that certain submissions (INDs/NDAs/BLAs etc.) include standardized tabulation data – specifically SDTM for clinical studies and SEND for applicable nonclinical studies (^[4]). The FDA’s Data Standards Catalog lists the required standards and effective dates. In brief, all in-scope nonclinical studies (e.g. general toxicology, carcinogenicity, safety pharmacology, and DART studies) “started on or after” the catalog’s deadlines must be submitted in SEND format (^[4]) (^[5]). If the specified standard is not used, an application may receive a refuse-to-file (RTF) action (^[4]). In addition, submission files must meet the technical rejection criteria laid out in the FDA’s Study Data Technical Conformance Guide (TCG) (^[4]) (^[7]).

SEND is a specialized implementation of the CDISC Study Data Tabulation Model (SDTM) for animal (nonclinical) data. In CDISC’s own words, “SENDIG is developed in reference to a specific SDTM model. … The SDTM is cumulative – each new release builds on the previous model. Therefore, the models are backward compatible” (^[3]). In effect, SENDIG versions align with successive SDTM releases. For example, SENDIG v3.0 (June 2011) was designed to be used with SDTM v1.2 (^[1]), while SENDIG v3.1 (June 2016) is based on SDTM v1.5 (^[2]). By CDISC policy, “Each new SEND release must include an assessment and realignment with the SDTM model and the SEND standard” to maintain cross-domain alignment (^[11]) (^[12]). In practice, this means new SDTM features (e.g. new variables or conventions) flow into subsequent SENDIG releases, ensuring that SEND always remains an “SDTM-like” IG for nonclinical data.

SEND was developed collaboratively by CDISC, the FDA, and industry stakeholders beginning around 2007–2011. The first published SEND Implementation Guide (SENDIG v3.0) appeared on June 17, 2011 (^[13]). It covered single-dose and repeat-dose general toxicology and carcinogenicity studies. In the years since, new domains and IGs have been added: for example, SENDIG v3.1 added dedicated Safety Pharmacology domains (Cardiovascular and Respiratory), and updated vital signs and microscopic findings domains (^[14]). Specialized IGs have also been released for unique study types. SENDIG-DART v1.2 (2023) covers Developmental and Reproductive Toxicology (EFD and juvenile studies), and SENDIG-AR v1.0 supports “Animal Rule” studies (where human trials are not ethical) (^[15]). Additional IGs are planned (e.g. SENDIG-Genetox for genetic toxicology studies). In tandem, controlled terminology and conformance rules have been developed to ensure consistent data use. CDISC notes that controlled terms are used for fields like specimen and finding identifiers, and conformance rule sets codify the SENDIG’s logic to support validation (^[15]).

In sum, SEND provides a rigorous, machine-readable tabulation of nonclinical study data, analogous to how SDTM provides standardized formats for clinical trials (^[16]) (^[3]). By submitting nonclinical datasets in SEND format with accompanying metadata (Define-XML) and documentation (nSDRG), sponsors enable regulatory review with powerful data analysis tools. As one industry author observes, standardized SEND data “allows regulatory agencies to streamline the review of the non-clinical sections of drug submissions” (^[17]).Sponsors also see benefits: CONSISTENT formatting and controlled vocabularies mean that data from multiple studies (even from different CROs or from prior legacy studies) can be merged and queried easily, facilitating meta-analyses and historical control comparisons (^[18]) (^[8]). In this report, we detail the evolution, structure, regulatory requirements, implementation practices, and future directions of SEND data standards for FDA nonclinical submissions.

Regulatory Requirements and Timeline

FDA Standards Mandate

The FDA issued final guidance clarifying that submission of standardized study data would become mandatory under FDASIA 745A(a) for certain application types (^[19]) (^[4]). Specifically, the FDA Data Standards Catalog (maintained by CDER/CBER/CDRH/CDV) lists which study types and effective dates apply. For nonclinical (GLP/NGS) studies, CDER had required SEND as of Dec 17, 2016 for NDA/BLA/IND submissions, and these rules have since expanded and been updated as new SEND IGs have been released. As of early 2023, the Catalog states:

• SEND IG v3.1.1 (the latest base SEND standard) became required for all in-scope studies with start dates on or after March 15, 2023 (^[5]).
• SEND datasets (v3.1 or v3.1.1) are now required for submissions to the FDA’s Center for Biologics (CBER) for studies starting after 3/15/2023 (^[5]). (Prior to that date, only CDER had enforced SEND.)
• The SENDIG-AR (Animal Rule) v1.0 is required for Animal Rule studies: effective 3/15/2022 for NDA submissions and 3/15/2023 for INDs (CDER) (^[20]).
• The SENDIG-DART (DART v1.1) is required for embryo-fetal development (EFD) studies: 3/15/2023 for NDAs and 3/15/2024 for INDs (^[21]).
• Define-XML v2.1 (metadata spec) is required for studies starting after 3/15/2023 (i.e. all new SEND submissions must use Define-XML 2.1) (^[22]).
• The SENDIG-Genetox v1.0 (for in vivo genetic toxicology) was added to the Catalog with support starting Dec 13, 2023 and requirement for studies on/after 3/15/2025 (^[23]).
• Additional requirements: In April 2024, CBER also began supporting and later requiring SENDIG-AR for CBER submissions (support started 3/26/2024; requirement begins with studies starting after 3/15/2027) (^[24]).

Table 1 below summarizes selected FDA deadlines and scope for SEND standards. (These dates come from the FDA Data Standards Catalog as updated March–April 2024 (^[5]) (^[20]).)

Table 1. Key FDA data standards requirements for nonclinical studies (from FDA Data Standards Catalog).

As the above shows, the FDA treats SEND very similarly to SDTM: once a standard is “required,” every new study in scope must provide data in that format. The FDA is strict about enforcement: submissions lacking a full SEND dataset (covering all endpoints modeled in the SENDIG for that study) plus the required define.xml and nSDRG will be rejected (^[4]) (^[7]). Indeed, the FDA clarifies that all in-scope studies that started on/after the effective date must use SEND or face a Refuse-to-File (RTF) decision (^[4]). For legacy nonclinical studies that began before the deadlines, sponsors may submit an abbreviated SEND package (TS dataset) with their original reports; in these cases the FDA allows an “abbreviated TS” in place of full datasets (^[25]).

In addition to the standards themselves, the FDA enforces separate business/validator rules and technical criteria on SEND data. The Study Data Technical Conformance Guide (TCG) contains business rules (e.g. v1.5) and validator rules (v1.6) that apply to both SDTM and SEND datasets (^[26]). For example, FDA’s Business Rules v1.5 (May 2019) document states it “applies to SDTM formatted clinical studies and SEND formatted non-clinical studies” (^[26]). These rules check conformance of variable names, controlledTerminology, dates, and relationships across datasets (^[26]). FDA staff and working groups (e.g. PhUSE) have also published additional clarifications – for instance, templates for the nSDRG have been circulated to help sponsors meet expectations (^[7]).

Regulatory Documents and Guidance

Key FDA guidances and resources include the Providing Regulatory Submissions in Electronic Format – Standardized Study Data guidance (Final, June 2021) and the Study Data TCG (^[19]) (^[27]). These outline the overall requirement under FD&C Act Section 745A(a) to submit standardized data, and specify that after 24 months from the final guidance, certain submission types must include machine-readable data. The Data Standards Catalog itself serves as an updated “standard operating procedure” for deadlines and is revised weekly. FDA also maintains quick-links pages for study data standards (^[28]). Notably, as of 2025 the FDA and CDISC are evaluating a future move away from SAS XPT files to a CDISC Dataset-JSON format for submissions, indicating ongoing modernization (^[29]).

Technical Structure of SEND Datasets

SEND data are delivered as SAS XPT (transport) files, one file per dataset (domain). The fundamental domains are analogous to SDTM’s, but tailored to animal studies. Key core domains in every SEND package include Trial Summary (TS), Trial Arms (TA), Trial Elements (TE), Trial Sets (TX), Demographics (DM), Disposition (DS), Subject Elements (SE), and Exposure (EX) (^[6]). The variables in these domains (for example, USUBJID, STUDYID, domain-specific keys) largely mirror SDTM conventions, but there are also new variables and domains as needed for study design. For example, the Trial Arms (TA) and Trial Elements (TE) domains define the planned experimental arms and treatment sequences – concepts with no direct analogue in clinical SDTM (which typically has ARM* datasets rather than TA/TE) (^[6]).

Each study’s SEND package also includes one or more findings/observations domains appropriate to that study’s endpoints. For a general toxicology study, such endpoint domains might include Body Weights (BW), Clinical Observations (CL), Laboratory Tests (LB), Microscopic Findings (MI), etc. For example, in a simple repeat-dose rat study the SEND package might contain TS, TA, TE, TX, DM, DS, SE, EX, plus BW.xpt (for serial weights), CL.xpt (clinical signs), perhaps LB.xpt (clinical pathology), etc. (^[6]) (^[30]). In practice, the exact list of endpoints is study-specific and guided by the SENDIG’s CoDEx mapping. Table 2 (below) illustrates a sample SEND file list and domains for a hypothetical toxicology study with body‐weight and observation data (^[25]).

Table 2. Example SEND datasets for a nonclinical toxicology study (as per FDA example) (^[25]).

SEND datasets are accompanied by Define-XML (version 2.1) metadata. This XML file describes all variables, labels, and controlled terms used in the XPT datasets, enabling FDA reviewers to navigate the datasets easily. Since March 2023 the FDA requires Define-XML v2.1 for new submissions (^[22]). (Define-XML v2.1 allows multiple standard references and new constructs for empty datasets, but for SEND it is largely similar to v2.0 with incremental features.) The related SAS XPT files themselves follow the XPT v5 format, which is the longstanding transport format accepted by the FDA (at least until conversion to JSON is implemented in future).

Another required component is the Nonclinical Study Data Reviewer’s Guide (nSDRG). This is a PDF document written by the sponsor that explains the SEND datasets in relation to the study report. The FDA’s Technical Conformance Guide states that the nSDRG is “recommended as an integral part of a standards-compliant study data submission.” (^[7]). While the exact format is not mandated, it typically includes explanations of any nonstandard domains or variables, definition of analysis flags, handling of blanks or missing values, and any assumptions or derivations. Industry working groups (e.g. PhUSE Nonclinical WG) have published templates to guide nSDRG content. Thus, each SEND submission is not just raw data; it includes the XPT files, Define-XML, an XSL stylesheet (for human viewing of the define.xml), and an nSDRG, all packaged in the eCTD (module 4) of the application (^[25]) (^[7]).

Domain Definitions and Controlled Terminology

The SENDIG specifies a set of about 30+ domains that cover standard study data (Table 2 lists some core ones). In addition to the core above, specialized domains exist for pathology (PA), ECG (EG), Ames test (MC), and others. Each domain has a set of variables, many of which are identical or analogous to SDTM variables (e.g. USUBJID, STUDYID, DY, SPECIES, SEX). Many variables in SEND are inherited from SDTM (e.g. the DM variables), and others are new or redefined for animal context (e.g. TESTART for test article, ACNAT for anatomical compartment). Users must populate variables with values from the standardized Controlled Terminology (CDISC CT). For example, permissible values for SPECIES are drawn from the SEND CT lists (e.g. “Dog” or “Rat” identifiers) published by CDISC. Versions of CDISC Controlled Terminology are updated with each SEND release (e.g. new specimen terms for DART studies).

The use of controlled vocabularies is a key feature. For instance, in the SENDIG v3.1 standard a controlled term variable MISTRESC (Microscopic Finding Result, Char) was defined in the MI domain. A recent analysis showed that in submitted SEND v3.1 datasets, 94% of all microscopic finding records had the MISTRESC field populated with a term from the official CT list (^[10]). (By contrast, v3.0 had no defined terms for this field, so 0% of records were coded.) Controlled terms thus ensure that results can be reliably searched and aggregated across studies (^[8]) (^[10]).

SEND also incorporates Supplemental Qualifiers (SUPP) domains (e.g. SUPPMI) for more detailed data. For example, if a rat had multiple microscopic lesion findings for one slide, additional findings may be captured in SUPP or related records. The Implementation Guide provides rules for how to link records across domains using standard keys (e.g. EXTRT, EXDOSE in EX domain link to TA/TE). There are also controlled timing variables (e.g. DY, DTC for study day or date-time) to standardize time points. Overall, the SENDIG enforces a strict tabulated model – unstructured text is not allowed in the main datasets, and free-format narratives should be kept in the CSR text (although some study-level comments can go in Comments variables). This rigor ensures the datasets are fully machine-readable.

Implementation and Data Quality Processes

Implementing SEND in practice involves generating the required domains and ensuring all rules are met. Many sponsors work with CROs or standards vendors to extract data from study databases or legacy records and convert them to SEND XPTs. A typical workflow includes: mapping raw data (e.g. animal tables, assay outputs) to the SEND domains; applying controlled terms; populating any study-level (TS) parameters; and generating Define-XML and the nSDRG. Specialized software tools or ETL scripts (SAS, R, or commercial tools) are often used. For example, the FDA has provided a sample SEND Starter Package and DEFINE.XML-XML Schema for guidance.

Quality control is critical. Because the FDA will technically “validate” the submission, sponsors must run conformance checks beforehand. CDISC publishes SEND Conformance Rules (separate from FDA’s Business Rules) which codify many SENDIG rules. These can be checked with validation tools (e.g. Pinnacle 21, SAS CRT). The FDA’s own Business Rules (v1.5) and Validator Rules add FDA-specific checks (e.g. required columns, code lists). Indeed, FDA reviewers have noted common errors in submitted SEND data, such as missing records or mis-spelled terminology. The agency even offered webinars on “common issues” when reviewing safety pharmacology SEND data (^[31]). These errors can jeopardize a submission: if a dataset has structural or content errors (per Appendix F of the TCG), the entire submission can be rejected. Sponsors therefore perform careful QC, often involving review of the nSDRG by nonclinical scientists to ensure alignment with the source study reports.

The Nonclinical Study Data Reviewer’s Guide (nSDRG) deserves special mention. While not strictly enforced, it is “recommended as an integral part” by the FDA (^[7]). In practice, the nSDRG is often the hardest document to prepare, because it requires explaining any deviations from pure SEND. For example, if a company adds a custom domain or leaves a SEND domain empty (because a particular endpoint was not measured), the nSDRG must note and justify this. A typical nSDRG will describe each domain file, define any custom columns, list the tables and listings produced by each dataset, and specify the controlled terminology versions used. FDA has provided templates (PhUSE has an nSDRG template) to ensure consistent information. In short, a SEND submission is more than just data files: it is a structured data package complete with metadata and documentation.

Analysis and Case Examples

One of the major motivations for SEND is to enable analysis of aggregated toxicology data. With standardized formats, the FDA can build repositories (such as the FDA’s open Janus database) where nonclinical datasets from multiple studies are pooled. Researchers and reviewers can then query these data across studies and sponsors. A striking example comes from a cross-industry research report (Carfagna et al. 2020) that analyzed SEND data in the FDA repository (^[8]) (^[10]). In that study, analysts posed queries about nonclinical findings (e.g. “What are the most common microscopic findings in male beagle dogs age 12–18 months dosed by oral gavage?”) and succeeded in retrieving answers by leveraging SEND domain data. They found that in control dogs, findings such as mature testis lesions, pituitary changes, and thymus alterations were among the most frequent – consistent with known background effects (^[32]). Crucially, they noted that the introduction of controlled terms in SEND v3.1 made this analysis possible: “Perhaps the most significant improvement facilitating cross-study analysis was the introduction of Controlled Terminology in SENDIG v3.1 for microscopic findings” (^[8]). Indeed, their Table 6 shows 3,802,024 total microscopic finding records in the query (from thousands of animals across studies) (^[33]), of which 487,216 records (94%) had standardized ‘MISTRESC’ terms (^[10]). Without the controlled terminology, automated queries would have failed or required manual coding.

These kinds of case studies illustrate the power of SEND data aggregation. They also provide quantitative insight: for instance, Carfagna et al. reported that 3,543,911 out of 3,802,024 records (98%) had valid animal age information, demonstrating that key metadata are usually present (^[34]). Analysis of such data can reveal incidence of rare findings, help set historical control ranges, or even support AI modeling. From the sponsor’s perspective, submitting good quality SEND data can therefore yield feedback and insights long after submission. In discussions, regulators have pointed to ongoing research leveraging SEND data for predictive toxicology and cross-study safety analysis.

In practice, sponsors have reported case-study successes. For example, large pharmaceutical companies have internally used SEND to consolidate data from legacy studies and to prepare integrated analyses. Converting a legacy distribution toxicology study into SEND can be like charting a process; one consultant notes that converting older studies (pre-2016) is tedious but valuable, because it enables integration into overall safety summaries and helps avoid review delays (^[35]). Moreover, SEND data are beginning to be used by sponsors to generate ADaM-like analysis datasets for toxicology (though ADaM is not currently required for nonclinical), pushing data analysis beyond regulatory compliance into research uses.

Implications, Benefits, and Challenges

Benefits: The uniform SEND standard brings multiple advantages. Regulatory reviewers no longer have to manually parse PDF tables; instead they can load the datasets and automatically run safety analyses (as was done in the Carfagna study). This makes reviews faster and potentially more thorough. For sponsors, having structured data upfront can simplify later life-cycle uses (e.g. additional analyses, regulatory reporting for post-approval studies). Standardization also facilitates data sharing between companies and with consortia (for example, aggregated SEND datasets are increasingly used in toxicogenomics and drug repurposing research). In sum, experts echo that SEND “improves the quality of the data” and “reduces the cost of implementing the standard” in the long run (^[36]). In fact, industry sources note that submitting standardized data can “increase the likelihood of a successful review at the first attempt” by eliminating workflow bottlenecks and providing clear, consistent data to reviewers (^[18]).

Challenges: On the other hand, implementing SEND is nontrivial. Organizations must build or license tools to convert legacy data, map to new domains, and perform extensive QA. There is a learning curve: toxicologists, data managers, and programmers must all understand the SENDIG rules. Sponsors must also manage versioning (e.g. choosing v3.1 vs 3.1.1). In some cases, not all collected endpoints fit neatly into existing domains, leading to creation of custom domains. (SEND v3.1 added flexibility to define custom domains.) In rare cases, ambiguous data or missing information from old studies must be handled carefully; the nSDRG becomes critical here. Failure in internal QC can lead to submission delays or FDA queries. The FDA’s recent webinars on common SEND issues (^[31]) underscore that even well-prepared sponsors sometimes submit incomplete or inconsistent datasets (missing records, incorrect timing, wrong XT formats, etc.).

It is also an organizational change. Pharmaceutical R&D groups have traditionally focused on narrative reports for GLP studies; transitioning to a data-centric approach requires process changes. Smaller companies or biologics firms (which faced SEND mandates only in 2023) may struggle more than large pharmas. Globally, the FDA is ahead: as of 2026, agencies like EMA or PMDA do not strictly require SEND (though ICH PHS harmonization encourages CDISC standards, and discussions about nonclinical data standards are ongoing internationally). Nonetheless, experience with the FDA indicates that SEND becomes a permanent part of the nonclinical documentation ecosystem.

Future Directions

SEND standards continue to evolve. CDISC is planning SEND v4.0 (expected Q4 2026) which will incorporate new domains like Immunogenicity Specimen Assessments (IS) for antibody data (^[37]). This reflects feedback from CBER on including ADA (anti-drug antibody) data. The FDA is already providing interim guidance: until IS exists, CBER suggests placing immunogenicity results in the LB domain or a custom IS domain (^[37]). Similarly, updates to existing domains (e.g. for new assays or advanced assays) will come as the science advances.

Other technological trends are also shaping SEND’s future. As noted, FDA/CDER are exploring CDISC’s Dataset-JSON format to replace XPT (^[29]), which could modernize how Send data are packaged. Machine learning and AI from big data are another frontier: aggregated SEND repositories may feed models that predict toxicity. There are also efforts to streamline cross-center data flow – for example, using standardized SEND data in the agency’s internal systems (FITS, Janus) or for sharing with external reviewers.

On the regulatory side, enforcement remains a focus. The TCG and business rules continue to be updated, so sponsors must stay current. The FDA has committed to publishing support and end-dates for older SEND versions, pushing industry to migrate to new versions. Meanwhile, the culture of clinical–nonclinical data integration is growing: with both arms of drug development using CDISC standards (SDTM/ADaM and SEND), cross-domain analyses (e.g. linking PK from animals to humans) become more feasible. This “fit-for-use” principle of standards (^[36]) will likely encourage SEND to become a foundational asset not just for submission but for R&D analytics.

Finally, training and community support (PhUSE CONs, CDISC tutorials, etc.) will expand. The FDA’s outreach (webinars, KickStart/OCS services) and industry best practices (SAS tools, open-source scripts) mean that over time the barriers to SEND compliance will lessen. Academic research may also leverage SEND: journals may begin expecting supplemental SEND datasets alongside preclinical study publications, paralleling clinical trial data initiatives. In all, the move to SEND is a step in the digital transformation of drug development. Its trajectory suggests deeper integration of data standards into both regulatory review and scientific innovation.

Conclusion

The SEND (SDTM-like) Implementation Guide has rapidly become the required standard for FDA nonclinical submissions. From its origins in 2011 through the latest 2021 and future releases, SEND has evolved to capture the complexity of animal studies in a structured, harmonized way (^[1]) (^[17]). The FDA now enforces SEND for virtually all nonclinical efficacy/toxicology studies in new INDs/NDAs, reflecting a regulatory commitment to data standardization (^[4]) (^[5]). This alignment between SEND and SDTM ensures that nonclinical and clinical data can be jointly understood and analyzed.

The practical impact is significant: sponsors who invest in SEND implementation benefit from more efficient FDA reviews and the ability to run powerful analyses across large datasets (^[6]) (^[8]). The FDA and CDISC’s continued collaboration – through controlled vocabularies, versioning, and new tool support – helps make SEND a living standard that keeps pace with science. At the same time, sponsors must contend with the challenges of converting legacy studies and meeting strict conformance. However, well-implemented SEND data can avoid RTFs and can even yield business intelligence (e.g. safety meta-analyses) that accelerates drug development.

Looking forward, SEND will likely expand (e.g. new domains in v4.0) and integrate with modern data ecosystems (JSON, analytics platforms). The foundational principle remains that standardized nonclinical data benefit all stakeholders – sponsors, regulators, and ultimately patients – by making safety data more transparent, comparable, and usable (^[18]) (^[8]). In conclusion, implementing the SEND “SDTM-like” standard is now an essential part of nonclinical research and regulatory strategy. All claims and guidelines cited here are drawn from FDA guidances, CDISC documentation, and published studies, and reflect the current state of knowledge as of 2025 (^[4]) (^[5]) (^[10]).

References: Registered eCitations from regulatory guidance, CDISC standards, and peer-reviewed literature as indicated (e.g. FDA Technical Guides, CDISC SENDIG releases (^[1]) (^[2]) (^[3]) (^[8]) (^[10])).