Nonclinical documentation in the CTD
The nonclinical documentation of the Common Technical Document is the nonclinical overview in section 2.4, the nonclinical written and tabulated summaries in section 2.6, and the nonclinical study reports in Module 4 from which they are written. Each has its own content to carry, and the overview and summaries are planned and written back from the reports.
Documents with nonclinical content
Nonclinical regulatory documentation divides into three major domains: summary documents for regulatory submissions, study reports with original data, and the various regulatory documents presented during the life cycle of a compound.
Nonclinical medical writers are involved across the modules of the electronic Common Technical Document: in Module 1, the investigator's brochure; in Module 2, the nonclinical overview at 2.4 and the nonclinical summaries at 2.6, running from 2.6.1 through 2.6.7; and in Module 4, the nonclinical reports, which are the source documents used to prepare the regulatory documents. Nonclinical writers typically engage with subject-matter experts from the early preclinical research stage through the life cycle.
Module 2 nonclinical documents
Along with the investigator's brochure, the Module 2 nonclinical documents are the only nonclinical documents with clear guidance about their content, which ICH M4S supplies. They comprise three written summaries for pharmacology, pharmacokinetics and toxicology at 2.6.2, 2.6.4 and 2.6.6, which present an integrated summary of findings, including a brief summary of the principal findings and a concise discussion and conclusion; three corresponding tabulated summaries at 2.6.3, 2.6.5 and 2.6.7, which present the data in tabulated form with no interpretation; the nonclinical introduction at 2.6.1, a small document of a few pages containing a brief summary of pharmaceutical structure, pharmacological properties and intended clinical use; and the nonclinical overview at 2.4, which provides an integrated assessment of the safety, pharmacokinetics and pharmacology data in the context of the proposed clinical trial or label. Module 2 documents are submitted with marketing approval applications and with American investigational new drug applications, which are mandatory for initiating clinical trials in the United States. Module 2 nonclinical documentation is largely focused around the investigational new drug application in the United States and, later, around a product licence, marketing authorisation or new drug application, and it covers most drug classes, from small molecules through to CAR T cells and gene therapy products. Similar, though less detailed, documents are required for clinical trial applications outside the United States, such as the nonclinical sections of the investigational medicinal product dossier in Europe and part 3 of the Australian clinical trial exemption application. Tables similar to the tabulated summaries are prepared for clinical trial applications in China.
Nonclinical study reports in Module 4
Most medical writers with regulatory document expertise focus on Module 5, the clinical study reports, while Module 4, the nonclinical study reports, includes valuable content that affects the documents in Module 5. Module 4's content falls into three main categories: pharmacology (how a drug works, how it may affect organ systems secondary to its main target, and how coadministration of drugs may affect how they work), pharmacokinetics (how a drug is absorbed, distributed, metabolised and excreted, and the potential for drug interactions) and toxicology (how a drug may cause adverse effects, including single-dose toxicity, repeat-dose toxicity, and the potential for genotoxicity and carcinogenicity).
Nonclinical study reports are included in Module 4 of marketing approval applications and investigational new drug applications; in Europe and other regions a summary of the nonclinical findings is usually sufficient for a clinical trial application, although agencies may request study reports during review and all cited study reports must be available on request. For a first-in-human clinical trial application, 20 to 50 reports will be prepared on average, and even more are usually needed for a marketing approval submission. They run anywhere from 5 to 3,000 pages, range from early pharmacology work to complex toxicity studies conducted under good laboratory practice (GLP), and may have been prepared in-house, at a contract research organisation, or by another pharmaceutical company in the case of purchased compounds. Complex studies, most notably GLP general toxicity studies, include several substudies, so-called phase studies, which are ultimately incorporated into the final report, and coordinating the preparation and delivery of the phase reports can be complicated because they often come from different departments or contract research organisations. A draft GLP toxicology report submitted to the US Food and Drug Administration (FDA) must contain a signed pathology phase report and must be followed within 120 days by the final report with a detailed list of changes.
Nonclinical writers also support some studies in the clinical domain, Module 5: in vitro and ex vivo pharmacokinetic evaluations with human biomaterials, and certain parts of clinical studies. They frequently support method validation and bioanalytical reports, because analyses for both nonclinical and clinical studies are often performed by the same bioanalytical laboratory, and biomarker analysis and population pharmacokinetics studies, which are often conducted by nonclinical or cross-functional units.
The turnover of nonclinical reports is high, because of the relatively short duration and large number of studies; reports are produced on an ongoing basis to inform regulatory and managerial decisions, with new findings sometimes leading to additional investigative studies, and writing activities must be coordinated between many internal departments and contract research organisations. Companies establish their own procedures for preparing reports, starting with the fundamental question of whether to prepare complete regulatory reports routinely for all studies or only when a submission requires them. The first approach may waste resources, because many compounds do not proceed into clinical development; with the second, submissions can become extremely challenging because of tight timelines, which can be made worse by having to prepare many reports, sometimes for studies completed years earlier. Keeping track of all the reports in the nonclinical package is an important activity for a nonclinical submission writer, because the reports build on each other to form the full picture of a compound's pharmacological, pharmacokinetic and toxicological properties. The key difference between clinical and nonclinical regulatory writing is the wider variety and greater quantity of nonclinical reports required during drug development.
Nonclinical guidelines focus primarily on scientific methods and development strategy, and GLP and the other guidelines provide only high-level requirements for study reports rather than detailed guidance on report structure. Unlike clinical reporting, which is guided by ICH E3, there is no single detailed reporting guideline for nonclinical reporting, which is not surprising given the variety of nonclinical studies. The quality and validation requirements for the studies and reports vary greatly, from strict GLP rules for pivotal safety evaluations to a more academic approach for pharmacology reports, for which some health authorities do not even require signatures. The internal reporting criteria must therefore be set appropriately, but even where requirements for report content and formatting are set, many reports, especially for pharmacology, contain no standard text and must be written from scratch.
SEND data
SEND, the Standard for the Exchange of Nonclinical Data, is a nonclinical variant of the Study Data Tabulation Model, and the FDA requires electronic submission of data in it for certain toxicology studies, which allows it to do its own interpretation of the results. Before that requirement, raw study data were archived and only submitted on request, and the shift led to an industry-wide scramble to prepare standardised study outputs and to reform processes to include the data files in submissions to the FDA. Because the SEND files are included in Module 4, nonclinical writers involved in assembling nonclinical packages may find themselves coordinating both the preparation of study reports and the delivery of the files.
SEND data are specifically for the FDA, and the European Medicines Agency (EMA) is thought not to accept them. Having a contract research organisation prepare a study in SEND format is quite expensive, reports quite often arrive without it, and the client then has to be told to supply it, which ends in delays. The data are a very long listing in a specific electronic table format, which the FDA loads into another programme to get the results out. Effects might be very difficult to find in them, and one experienced writer would not recommend them for a writer's own overview, because they are just a very long listing. A writer needs to be aware, though, that a submission to the FDA has to include them.
Contract research organisations asked to make SEND listings say that the listings are exempt from GLP, while the rest of the study has to be performed under GLP; the listings are barely quality-controlled, there can be mistakes in them, and the FDA still accepts them. The FDA requires source data in this form and the EMA does not, a difference with in part a legal background. One European assessor holds that the FDA does not trust anything a sponsor writes and works bottom-up, looking at the study reports first and then moving up, whereas European assessors trust what is written in the documents, and that SEND is an extension of that approach, letting the FDA reassess the data itself so that it cannot be accused of any failings. The same assessor holds that liability, which matters more in the United States than in Europe, is the most important reason. After so much effort spent harmonising the guidelines, SEND is a point on which the two regions do different things.
There is an initiative in Europe, at the EMA, to start looking at SEND. None of the assessors are keen on it, because it would mean a lot of extra work and a great deal of storage capacity.
Planning back from the modules to the source documents
The work begins with the end in mind. The writer should engage early with subject-matter experts and the team and align early with strategic partners, in particular the regulatory lead; the source documents should be mapped as early as possible, the timelines built from the bottom up on the content mapping and the interdependencies, and documents developed strategically with reusable content wherever possible. From the kickoff of an investigational new drug application to its submission takes roughly six months under optimal conditions, sometimes more and sometimes less. Going from a source document to a regulatory document calls for a clear distinction between scientific and regulatory writing.
Section 2.4 is typically built from brief summaries of the pharmacology, pharmacokinetics and toxicology written summaries, and 2.4 and 2.6.1 are typically the last documents written, with some exceptions. The source document for the pharmacology written summary at 2.6.2 can be section 4.1 of the investigator's brochure, and writing that brochure section with the module in mind reuses quite a bit of content. Section 2.6.4 draws on brochure section 4.2 together with the analytical reports, the pharmacokinetic reports and the pharmacokinetic modelling; safety margins sometimes go there and sometimes in 2.6.6, depending on the key messages and how the team wants to proceed. The modules also depend on one another, the written summaries at 2.6.2, 2.6.4 and 2.6.6 coming before the overall documents at 2.5 and 2.4.
Beginning with the end in mind means recognising what content can be reused from the investigator's brochure or the briefing book, so as to avoid "info obesity", redundancy across modules. Mapping also covers the rate-limiting documents: pharmacokinetic–pharmacodynamic modelling and simulation is usually one of the last documents to arrive, and the pharmacologist has to be reminded to provide it as early as possible. In one writer's experience, toxicology reports are typically one of the rate-limiting steps.
A document cannot be taken too far before the feedback that has to be incorporated into the modules arrives: agency feedback, advisory boards, and the governance steps an organisation goes through in preparing the investigational new drug application. Best practice is to map out what feedback is coming before authoring, so that the application is prepared more efficiently.
State of the source documents
The first questions about the source documents are how many there are and what stage they are at. A writer would prefer final reports, and that is often not what arrives. Summaries and draft reports take extra time, because all the documents have to be updated later; in some cases a client sends Excel sheets for pharmacology studies and asks for them to be used, and has to be told that a signed report is required.
In one contract research organisation's first-in-human packages for small molecules, the mean number of studies across pharmacology, ADME (absorption, distribution, metabolism and excretion) and toxicology was 37, which it uses as an average when quoting, because the number of studies sets the cost. Some companies sometimes have double that, and monoclonal antibodies tend to have about 50% fewer studies, because not much ADME or toxicology work tends to be done with them.
Study data for the United States have to be supplied in the SEND electronic format; this is not strictly a writing requirement, but a writer needs to be aware that the studies have to be in the correct format. Study reports from work with companies in Asia are quite often in another language and have to be translated, and the documents cannot be written while the translation is awaited. And data often do not arrive all at once: some come this week and some next month. Some companies also ask the writer of 2.4 and 2.6 to write the investigator's brochure, which helps because the data can be transferred between them; otherwise the timeline has to be tied to the external preparation of the brochure and, eventually, the submission.
Transparency about what is missing belongs to the assessment of the source documents too. A sponsor can send a package of study reports, say that nothing is missing, and only remember the absent reports once the reports supplied are found to refer to them. Regulators will pick up work that has been done and not disclosed, and finding it out is not good for the interaction.
More and more biotechs and pharmaceutical companies run their phase 1 first-in-human studies outside the United States and then file an investigational new drug application to enter phase 2 there. Where the phase 1 was run in China, the company probably has the investigator's brochure and other modules in Chinese, and has to decide whether to write the brochure and section 2.4 from scratch from all the data or to translate its documents and have them reworked into an acceptable form. The first step is a first-line assessment of whether the study reports and the data in them are acceptable to, say, the FDA, and quite often there are many problems: what is acceptable in one region, for example, is not acceptable in another. The programme of studies that has to be performed makes little difference whether the destination is Europe or America, because it is covered by the ICH guidelines, which Japan also has to go along with.
The nonclinical overview
The structure and content of Modules 2.4 and 2.6 are specified in the ICH M4S guidelines, with answers to the most common issues raised provided as a separate document. The guideline says that the nonclinical overview should generally not exceed about 30 pages. Overviews that one organisation had been involved in ranged from 8 to 35 pages. Section 2.4 needs to be concise and lean, and avoiding repetition of the written summaries in it takes being very critical and striking out text that repeats them. It may sometimes be important to get a toxicologist involved in 2.4, because they can strip out the extra material.
The nonclinical overview should provide an integrated overall analysis of the information in the CTD. The interpretation of the data, the clinical relevance of the findings, any association between nonclinical findings and quality aspects of the product, the results of clinical trials or effects seen with related products, and any implications of the nonclinical findings for the product's safety in humans should be addressed in the overview. The overview should discuss and justify the nonclinical testing strategy. It draws conclusions on safety for the intended clinical trial or therapeutic use. It should take account of the scientific literature and the characteristics of related products: a finding seen with a drug in the same class should be discussed. It should also describe impurities, provide references from the literature and list the available study reports. The overview is thus an integrated and critical assessment of the pharmacological, pharmacokinetic and toxicological aspects of the product in animals.
Section 2.4 is not a sales pitch: claiming that a molecule is better than everything else on the market does not go down well. It should state which regulatory and ICH guidelines were followed, and justify any deviation from the guidelines on the conduct of the studies. It should reference any regulatory agency interaction, such as a pre-IND meeting, which saves the regulator going back to find out whether a meeting occurred and what was discussed. It should say where the studies were performed and comment on their GLP status. China is not a member of the OECD family of GLP, so for toxicology studies conducted there the sponsor has to make sure with the regulators that they will accept them. Studies conducted in a country outside the OECD Mutual Acceptance of Data system may not be accepted by some health authorities, such as those of South Korea and the United Kingdom, which do not accept GLP studies conducted in China. Interpretation of the GLP guidelines can also differ between countries. Regulators expect a set of study types to be followed, so any study added to that set or omitted from it has to be explained.
Inadequate discussion of impurities or excipients often leads to delays in regulatory interaction, and a novel excipient not previously used in any formulation needs a discussion of what studies were performed on it, which may still not make it acceptable. Delays in development around impurities and degradants that were not properly covered and discussed are frequent, and new genotoxicity and toxicology studies on the impurities are sometimes needed.
The argumentation in the nonclinical documents changes as the compound advances in clinical development. Early submissions such as an American investigational new drug application focus on nonclinical evidence for safety and pharmacological activity, including justification for the chosen nonclinical programme and for the adequacy of the safety precautions set in the clinical protocol. For a marketing approval the writing task is different, because the nonclinical data must be discussed in the context of the available human data, with an emphasis on how they align with the effects observed in humans, such as pharmacological mechanism or adverse reactions, and with the labelling claims. The summary documents also become more extensive, because the number of studies to include generally grows after the first-in-human trials.
Subsections of the nonclinical overview
The nonclinical overview is divided into subsections with the following contents.
| Subheading of 2.4 | Contents |
|---|---|
| 2.4.1 Introduction | |
| 2.4.2 Overview of nonclinical testing and GLP statement | The nonclinical testing strategy should be discussed and justified. There should be a comment on the GLP status of the studies submitted. Any association between nonclinical findings and quality characteristics of the investigational compound, results of clinical trials, or effects seen with related products should be indicated as appropriate. |
| 2.4.3 Pharmacodynamics | Studies conducted to establish the pharmacodynamic effects, mode of action and potential side effects of the drug in development should be evaluated. |
| 2.4.4 Pharmacokinetics | The relevance of the analytical methods, pharmacokinetic models and derived parameters should be discussed. Interspecies comparisons of metabolism and systemic exposure should be discussed and comparisons between animals and humans made (AUC, Cmax and other appropriate parameters). The limitations and utility of the nonclinical studies for the prediction of potential adverse effects in humans should be discussed. |
| 2.4.5 Toxic effects | The overview should evaluate the time of onset, severity, duration, dose dependency and degree of reversibility of toxic effects in the context of differences related to species or gender. Genotoxicity, carcinogenic potential, and effects on embryo-fetal development and juvenile animals should be discussed. Extrapolation of the data from animals to humans should be considered in relation to the species and number of animals tested and issues concerning dosage and systemic exposure (no adverse effect levels, toxic doses, etc.). Recommendations should be made on exposures in humans. Logical and well-presented conclusions should be made on the implications of the nonclinical data for safe use of the drug in humans. |
Pharmacology, pharmacokinetics and toxicology in sections 2.4 and 2.6
For pharmacology, sections 2.4 and 2.6 need to present the appropriate models to show that the drug has a pharmacologically active, efficacious action, and to mention the dose levels. They need to explain what was done on secondary pharmacology, the search for off-target effects, and to discuss any indication of potential off-target effects; for a monoclonal antibody that means the tissue cross-reactivity study. Where a biomarker with clinical relevance was used, they need to discuss it and how it is relevant, a B-cell response with monoclonal antibodies being an example. For safety pharmacology, they need to explain the core battery of studies on the central nervous system, respiratory function and the cardiovascular system. Where a signal appears, in the cardiovascular work for example, further studies probably have to be done, and the documents need to record them and why they were done.
For pharmacokinetics and ADME, the documents need to discuss the methods used, such as LC-MS, and their sensitivity, but not in too much detail. Companies often supply pages of analytical methods that bore everybody, the regulators included, and they have to be persuaded to reduce those to a page or two. Units need to be consolidated across studies, because micrograms in the dog study and nanograms in the rat study leave a regulator to do the conversions themselves, which drives them mad. Any pharmacokinetic or pharmacokinetic–pharmacodynamic modelling relating animal exposure to expected clinical exposure needs to be discussed, with how it is relevant. Drug–drug interaction potential, a very popular area, needs to be discussed from induction, inhibition and transporter studies, with how relevant it is to patients, who will quite often be on other medications. Metabolites need to be discussed as the data become available through development: whether there are unique human metabolites or major metabolites, and how they were handled.
Toxicology starts with only general toxicology and genotoxicity studies, and reproductive, carcinogenicity and other studies are added as development proceeds. The species, number of animals, route of administration, dose levels and duration need to be described in an ordered way, and the real study findings highlighted: those relevant to human safety, and whether they are reversible. A finding that is a known class effect is worth referencing, so that the regulators know the sponsor knows it too. For example, an oligonucleotide classically produces histopathology changes in the liver and kidney that are not adverse and are due to the slow clearance of these large molecules. Regulators, especially in Europe, push back a lot on species choice. Using rats and dogs because they have always been used is not good enough: regulators want to know why each species was used, and a document using a primate has to say why a primate was used rather than a dog. With one molecule, monkeys were chosen because, compared with the dog, they had homology to the human sequences of the target. Where a drug crosses the blood-brain barrier into the central nervous system it has the potential for addiction, and the array of animal studies done for that has to be discussed; where a drug is going into a paediatric population, studies in juvenile animals are sometimes needed and have to be discussed in an ordered way.
The genotoxicity conclusion needs careful wording: "a battery of studies confirm no evidence of genotoxicity" is not the same as "no genotoxicity has been reported", which suggests to a regulator that no genotoxicity testing was done. Positive genotoxicity findings need proper interpretation: whether they are relevant, and whether they occur with similar compounds, tyrosine kinase inhibitors in oncology, for example, always being positive yet acceptable because they are going into cancer patients. The findings are related to those of any carcinogenicity studies in a weight-of-evidence approach, and all of it goes into section 2.4.
Nonclinical results in the label
For marketing approval submissions, nonclinical information is traditionally seen as corroborating clinical evidence: safety assessments can provide mechanistic explanations of adverse reactions observed in humans, and pharmacokinetic evaluations can inform decisions on clinical pharmacology studies, but it is the clinical trial results that actually support labelling claims. Some data cannot be intentionally obtained in clinical trials, most notably data on carcinogenicity and reproductive toxicity, so labelling is directly influenced by the results of those nonclinical investigations, along with some nonclinical pharmacokinetic data, such as drug–drug interaction or transporter studies.
The written and tabulated summaries
The main purpose of the written and tabulated summaries in Module 2.6 is to provide a comprehensive factual summary of the nonclinical information on pharmacology, pharmacokinetics and toxicology; the written summaries are generally in the region of 100 to 150 pages, and ICH M4S provides 34 templates for preparing the tabulated summaries. In one organisation's first-in-human investigational new drug applications, the toxicology written summary varied in length, and was sometimes only a few pages, usually for a very specific product such as a monoclonal antibody.
Section 2.6 should have no invented sections and no renumbered ones, and the location of data in it matters. Some human data raise the question whether they go into the clinical overview or the nonclinical one. Juvenile animal data may go into the general toxicology section, because it is a toxicology study in young animals, or into the reproductive toxicology section. Toxicokinetic data is a very common case: some companies put the exposure data from their toxicology studies into the pharmacokinetic written summaries and tabulations and some put it into toxicology. There is no hard and fast rule; one organisation's preference is to put it into the pharmacokinetic section and cross-reference it from toxicology.
The tabulated pharmacokinetic and toxicology summaries are at 2.6.5 and 2.6.7, while what appears in the pharmacology tables may be very different from compound to compound. Pivotal studies have to be performed with validated methods under GLP, and the validation also has to be tabulated in the modules. A non-pivotal study is quite easy to tabulate, with the species, the method of administration, the duration of treatment and a listing of the effects observed. At the end of a general toxicology study the animals are taken apart into about 40 to 50 tissues and organs, which are examined macroscopically or microscopically, and all of it has to be tabulated. The tables for a single 13-week rat study list the species, duration, animal age, exposure data such as AUC and Cmax, body weight and body weight gain, clinical observations during and after administration, haematology with its statistics on each of several occasions, clinical chemistry, immunophenotyping for a compound with an immune function, and the histopathology. The genotoxicity tables carry the study description, how often it was replicated, whether it was GLP-compliant, the positive controls, the cytotoxic effects that are not genotoxic effects but are needed for interpretation, and all the results, for an Ames test across all five salmonella strains. Reproductive toxicology tables cover both the dams and the offspring, including whether there was an effect and everything found in the skeletal and visceral examinations, and can run to pages of their own.
Contract laboratories sometimes supply tables in this form but not very often, so the writer very often builds them from the reports. Putting the tables together is a lot of work, and very good quality control has to be performed on the data submitted.
The guideline says that only noteworthy findings should be put into the tabulations, which leaves the question of what counts as noteworthy, and one organisation involves people with a toxicology background at this point.
Some companies have submitted early investigational new drug applications with no tabular summaries at all, and the FDA has accepted them. Contacts at the FDA said that it accepts this for first-in-human applications, but would like some tabulated summaries for submissions with a substantial number of studies, later in development.
The FDA's review
A writer spends a great deal of time on section 2.4, yet the first thing an FDA reviewer does is look at a study report: the FDA's review works bottom-up from the study reports. So putting everything into 2.6 sometimes does not matter.
Condensing and consistency
Cutting and pasting from study reports directly into sections 2.4 and 2.6 drives regulators to distraction, is instantly recognisable when reviewing the document, and ends with many further reviews. Condensing a toxicology report in which much of the detail is not drug-related is part of the skill, and it is hard for a writer. Writers with a nonclinical background know what to take out, and one experienced toxicologist holds that otherwise it is a matter of experience and of being properly mentored. It is really important to have someone oversee what has been written and perhaps strike out some of the unnecessary detail.
Inconsistencies across documents are to be avoided. In one case, a range-finding toxicology study in a monkey recorded an effect on blood pressure, which the study report described as pharmacologically relevant; section 2.6 repeated that, and section 2.4 did not mention it at all. The proper GLP toxicology study, in many more monkeys at the same dose level, did not see the effect, so the finding in the one animal was not in fact pharmacologically related and was coincidental, but the study report had said otherwise. The documents need to be consistent, and such a finding has to be addressed rather than left hanging.
Four rules sum up the work: get the documentation right; know the regulatory guidelines and requirements; do not hide problems, and be transparent; and get toxicology input, even if only on an overview of what has been written. A poor dossier ends in delay, which is delayed benefit to patients. The FDA has published areas in which it found nonclinical data in investigational new drug applications and new drug applications insufficient, and properly written documentation does not attract those issues.