Run quality control as a defined review
Quality control is a defined review that checks a document against its sources and against the standards it is written to, ideally by someone other than its author. How well it works turns on when it is run, who runs it, what the writer hands over with the document, the terms agreed before the first read, and how much time it is given.
What the check is
The guideline for good clinical practice of the International Conference on Harmonisation defines quality control as "the operational techniques and activities undertaken as part of the quality assurance system to verify that the requirements for the quality of trial-related activities have been fulfilled", and it should be applied at each stage of data handling — collection, storage, processing and reporting — so that all data are reliable. For medical writing, it generally means verifying that a deliverable, such as a clinical study report, a marketing application, a protocol or a manuscript, is accurate against its source documentation, such as the tables, figures and listings and the protocol, and conforms to the style guide and template where one is provided. The process is important to ensure that accurate and high-quality data are presented. Any medical writing that reports data needs some kind of quality control, so that documents are not riddled with errors, and it may take in consistency checks, editing and content verification. In regulatory writing, quality control is an integral part of making submissions accurate and consistent.
Quality control means checking a document to ensure the best possible quality of the output — its content, style and format. It is neither a check for typos on a rough first draft nor a quick skim of a near-final document, but a comprehensive and detailed review. That quality does not just happen but results from a systematic review. Readers all expect high-quality documents, and that requires an organised review process. A good review takes time but is well worth the effort.
The check takes in two kinds of work, verification of the content against its sources and a reading for consistency, style and language. One technical editor, who performs quality control on clinical study reports, new drug applications and integrated summaries of safety, describes its purpose as ensuring "a certain level of quality within a clinical document prior to delivery", and divides it into a content check, which verifies the scientific data by cross-checking the source documents against the working document, and a technical editing check of style, formatting, grammar, usage and spelling. The check is made against source data, by comparing data in tables and confirming that analytical statements are supported by the data presented, and for consistency in spelling, grammar, punctuation, formatting, style and cross-references; a quality control specialist may also be called a data integrity reviewer, copy editor or proofreader. One contract research organisation, PPD, counts the editing of regulatory documents such as clinical study reports, protocols and investigator's brochures as quality control, because its editors verify data and the consistency of the text against source documents as well as checking grammar, spelling and punctuation. Quality control can be defined as checking consistency against a standard, and in writing it is more specific than review. Leaving out each kind of review — data, subject-matter-expert or editorial — has different implications: data and subject-matter-expert reviews can be critical for a regulatory submission, while an editorial review is often what a document's appearance needs.
When the check is run
Practitioners differ on when the check is run. One guide places it before the reviewers see the first draft, as an integral part of producing the document rather than a one-off step between the first draft and the final one, because every version should be of the highest quality. One technical editor runs it three times — when the document is a shell, at first draft and at completion — and finds that the level of checking needed decreases as the document moves through the process. For one quality control reviewer, the review is the final step. Another guide has the check typically come at the end of the process, after the document has gone through several rounds of review by numerous reviewers. For one quality control specialist, it usually happens when the document is near-final or final, as the last check before the document goes to the client and then to its audience, such as a regulatory agency or a journal.
Why a document needs the check
A reader other than the author catches what the author no longer sees. A fresh pair of eyes can spot errors an author has overlooked; people compensate for missing or erroneous text as they read, which is why the review brings fresh eyes and a new perspective; and a writer who has looked at a document for too long stops seeing its errors. Two further reasons concern documents with several reviewers: a document carrying potentially conflicting comments from them needs a new and objective look, and someone has to make sure that all the comments were incorporated.
Even the best writers are not infallible. It is extremely easy to copy wrong data into a table from source tables running past a hundred pages, to make an analytical statement the data do not wholly support, or to leave the full stop off a sentence. Inconsistencies such as whether a space follows a symbol in many ways matter less than wrong data, but a document inconsistent in many small things looks messy and inattentive, which can draw attention away from its key messages. The findings are used by regulatory agencies, health professionals, patients and the public to evaluate a drug's efficacy and safety, and can inform later studies, so data need to be presented accurately and clearly and a safety net is needed to catch errors and inconsistencies; the quality control specialist's fresh and objective reading for consistency, correctness and clarity is that net.
Errors cost the writer as well as the science. Reviewers are customers whether they sit inside the writer's company or outside it: a high-quality document tells them they are working with an accomplished writer, and a poor-quality one can jeopardise repeat work and may limit job prospects. Mistakes can also make a reader question the validity of the content and may lead to errors of interpretation.
Errors also act outside the document. Quality control gives assurance that the data on which decisions are made, and the materials associated with a product, are complete and accurate and that patient safety is protected, and regulators may not believe a company's claim that a product is safe and effective if its marketing application is riddled with errors. The Food and Drug Administration has indicated that errors in datasets, such as transcription or transposition mistakes, can reduce confidence in the overall accuracy of the data and may support a refusal-to-file decision, under which the application is not reviewed until the problems are fixed. Labelling errors have led to recalls: in 2014 one manufacturer voluntarily recalled a ketorolac tromethamine injection labelled with an expiry date longer than the product's known stability, and another recalled a highly concentrated potassium chloride injection whose overpouch was mislabelled with a higher concentration, which could have had serious and life-threatening consequences in a high-risk population given a lower dose than intended.
Simple omissions can be funny — morality rates for mortality rates — or dangerous, 10 g for 10 mg, and mistakes in documentation are propagated further by the copy-and-paste practice that is the norm in regulatory documents. Writers normally do not deal directly with study sites or participants, but they have the power of words, which comes with a huge responsibility: a small typo in documentation can have huge consequences, and research shows that many safety incidents in healthcare may be attributed to communication errors.
Who performs the check and what they need to know
The review should ideally be done by someone other than the author, and most often that is another writer in the same company, of less, equal or greater experience. A writer working alone may have to check their own work, and the principles are unchanged. Where a document has been reviewed and needs quality control on a later version, the same reviewer has the advantage of already knowing the document, and some companies employ people whose main role is performing these reviews, an asset where available.
An effective reviewer should have attention to detail; analytical ability; problem-solving ability, such as identifying the correct source document when it is not clearly indicated; organisational skills; interpersonal skills and the ability to communicate clearly; editorial and proofreading ability; knowledge of medical writing processes, including how team members influence a document's content; and strong computer skills with common applications such as Microsoft Word. The same skills make a better writer, and a writer may be asked to perform the check directly by a client or for a colleague before a document goes to a client; having them lets a writer put forward a high-quality document from the beginning of the process. Developing them as a writer can lead to higher-quality documents earlier in the process and greater client satisfaction.
A specialist can do much more than proofread for spelling, grammar and punctuation or check decimal places. Specialists understand complex medical and statistical terms, know what it means for a study to be underpowered or for p values to be inappropriate, and can therefore say when an analytical statement might be too strong given the study's methodology as well as its data. They can identify the elements of a well-written manuscript and how a study's salient messages are carried in a discussion, may know more than a writer new to manuscripts about a journal's submission requirements and its submission portal, and, for a writer who has only recently entered the profession, can be a resource for regulatory guidance and where to find its latest updates; under-using their abilities would be a waste.
The quality control team is responsible for maintaining the integrity of materials while keeping to the allotted timelines and budget, and toward the end of a document's course the quality controller often meets its roadblocks. In a 2018 survey of 11 quality controllers, the tasks respondents predominantly named as the job were reviewing grammar, spelling and punctuation, checking data, confirming adherence to format and style, and copyediting, and the person doing it was described as often expected to take on "other duties as assigned". Of the respondents, 72.7% worked mostly on regulatory documents, 36.4% on manuscripts and peer-reviewed publications, and 27.3% on educational, marketing or public relations materials; most worked in pharmaceutical or biotechnology companies (44.4%) or freelance businesses (33.3%), and they had worked in quality control for an average of 3.9 years.
Training in how to perform a good review always benefits whoever does the check. Where none is available at work, it can be had from EMWA's Editing and Proofreading Essentials workshops, parts 1 and 2, at the foundation level of its professional development programme, and from courses run by bodies such as the Society for Editors and Proofreaders and the DIA, some of them online; a course gives a new writer or reviewer a good start, and an experienced writer can still learn something new from one. Learning ways to make sure that all mistakes are caught can help anyone working in quality control. Training options that help new hires and seasoned reviewers alike fall into three categories: collaborative, self-study essentials, and self-study over time. One trainer recommends taking lessons learned, an option of self-study over time, by meeting as a team and going over the processes together, and stresses curiosity, building the skill set through things done before and after a review. Continuing to ask for quality control feedback, and learning from best practices, helps a writer gain perspective on the process.
What the check verifies
The content half of the check reconciles a document with itself and with its evidence. For a manuscript these are internal fidelity, or precision, which is consistency across its sections, and external fidelity, or accuracy, which is compatibility with the published references and the study data. No credible manuscript contains self-evident internal inconsistencies, or purported statements of fact that a single search can disprove.
Internally, findings cited in the abstract must match those in the results text, which in turn match those in the tables and figures. Every result should be motivated in the methods: no data should appear for the first time in the results, and no result of an assessment the methods specify should be left out. Findings, numbers above all, need to be both minutely correct against their sources and intuitively sound. A series of percentages that does not sum to 100 because of rounding, or because missing data left different denominators for different assessments, should carry a footnote saying so. A final reckoning asks whether any endpoint stated in the methods goes unreported in the results, or the reverse. Externally, reported data must minutely match their external sources, including the raw outputs — tables, figures and patient listings — provided by statisticians and the published references cited; reformatting tables and figures from the raw data rather than retyping them helps keep them consistent, because rekeying can introduce errors.
Common quality control checks fall into three kinds. Consistency covers internal consistency among sections, external consistency with a summary-of-changes document, and consistent formatting of headers, footers and captions. Editing covers spelling, grammar and punctuation; adherence to the style guide and template; logical and understandable flow; and abbreviations defined and used consistently. Content verification covers numbers and text aligned with the source; in-text tables and figures checked against their sources; references appropriate and accurate; and key messages aligned with the data presented.
Common errors are worth knowing for the concepts behind them, which can show up in other ways in any document: be wary of absolute statements and look for the exceptions, watch for missing units and information, and turn vague statements into specific ones. The grammar and punctuation errors and inconsistencies that one editor notices most often include commas, hyphens, colons and abbreviations. Knowing how the document was built helps the reviewer think like its writer — from a template, from an approved document of the same type in a different therapeutic area, or from a global module with country-specific updates — and each has its own things to watch for: template text that contradicts the de novo text, wrong key terms, the therapeutic area among them, and dosing and endpoints that can vary by country.
What the writer hands over with the document
Early in a career the check can feel like someone holding a magnifying glass to a carefully crafted and deeply loved document in order to pick holes in it. Most experienced writers love it, and in one editor's view a writer gains the most from a well-executed check only by understanding the best way to work with the specialist, getting as much out as they put in.
The frustration runs both ways: a writer can be frustrated by comments on consistency or wording they did not want checked, or by the absence of the minor fixes they expected, while a specialist can be equally frustrated by a writer who has not said what to check or how findings should be addressed. Five stages, which one specialist with experience as both writer and specialist has found useful in improving the efficiency and effectiveness of the check, can enable a writer to get the most out of a specialist.
- Proofread the document before sending it, which sounds obvious but is essential. The check is the final one before the document goes to the client, so a document carrying many consistency and data issues may need checking again once the comments are addressed, which adds time to a project that may already be time-critical and uses more resources. A document that has not been proofread can come back carrying so many comments and tracked changes that the crucial corrections — whether the protocol's endpoints are addressed in the results, whether the conclusions on efficacy and safety are supported by the data — are hard to see. Consistency software can speed the proofread up and help find issues that would take longer to find by reading, but it does not replace the writer's own proofread, because it will not find issues such as those with analytical statements or incorrect data.
- Allow a realistic amount of time: a consistency and data check on a 100-page clinical study report cannot be done in one day, a good specialist will say so at the outset or as soon as the work proves longer than expected, and a specialist will do what can be done for a quick turnaround as long as the quality of the check does not suffer.
- Be clear at the start about what is to be checked and how findings should be addressed, and never presume the specialist knows either, while the specialist always asks what needs checking and how the findings are to be addressed; a specialist told to skip something already checked still raises it with the writer where the later draft turns out to carry substantial changes, such as a lot of new data or analytical statements.
- Signpost the source documents: say where they are, whether there is a system for labelling them and whether a style guide applies, which saves the specialist searching and saves the writer answering repeated queries.
- Acknowledge receipt of the document and give feedback on what was good or helpful and what did not work, because that is how a specialist develops, and because people who feel valued become more engaged, come to work more often, are committed and feel accountable for their performance, so feedback can be a very powerful tool for improving the check.
The reviewer needs two things from the writer, where the sources are and what the check is to cover. The writer provides a narrative map indicating where the sources are, to help the reviewer perform the cross-checks; gives the location of all the source documents and states the scope of review expected — all of it, or specified text, sections, figures or tables, and whether it covers grammar, style and consistency; and, in a request for the check, states the type of review, the sections where only parts need checking, the file name and location, and the due date and time.
The review is not to be taken personally, because document preparation is collaborative. Before sending a document, the writer plans their schedule so as to be available for questions, gives the reviewer advance notice so they can prepare, and discusses complex study designs or data presentations with them, which helps reduce both the comments and the review time; knowing the style guide and keeping the language consistent helps reduce the comments further. Only the final draft is sent, because even a small change in a table can ripple through the document and force the review to be repeated.
The writer should be proactive in communicating this basic information. The writer also helps define the roles and expectations in the check, a subject-matter expert reviewing content rather than editorial aspects. Common document inconsistencies include documents not aligned with their sources, inconsistent terminology and style conventions, and errors of grammar, punctuation and spelling, and issues that may arise later in the check include too little time for it, vague expectations or instructions, and misplaced expectations of each role or review type. Identifying the standards that govern the document is a crucial step. Four questions are worth asking before the check to make sure the document is consistent: whether the writing aligns with its sources, whether it is easy to navigate, whether the same content is written about in the same way, and whether the same components look the same. Sources are identified clearly, with document identifiers such as the study identification, version number or date, and kept organised. A style guide can help keep the document uniform where many writing conventions are acceptable, specifying for instance the company and drug names, the preferred template and toolbar, abbreviations and terminology, and usage such as patient against subject. Consistent tools, a clear starting plan and concrete communication let a writer produce high-quality documents efficiently.
Terms agreed before the first read
Deliverables vary widely in style and convention, but these best practices, in order, make for an effective check whatever the document type.
- Establish expectations with the author: the type of check requested, how thorough it should be, what to do with questions, how and when the author wants to be contacted — as the reviewer goes, for example about unverifiable data, or in one consolidated list at the end — and how findings are to be documented, whether as tracked changes, comments or a spreadsheet of issues. Some documents have no source documentation, such as a protocol that has evolved past its concept sheet or synopsis, or have already been checked by functional experts, and then copyediting without content verification may be all that is needed.
- Identify any special circumstances, such as checking several related documents as part of a team — modules of a marketing application, say — where cross-references have to be verified with other members of the team; scheduling time with the author to go through them can help.
- Identify the standard operating procedures that apply: clients may have their own procedures, checklists and verification forms, and where a client's procedure conflicts with the reviewer's company's, confirm with the author which is to be followed.
- Confirm the timelines, including whether the due date ends the reviewer's involvement or follow-up questions from the author should be expected.
- Confirm the type of document and its source documents — familiarity with the document type tells the reviewer what sources to expect, and unfamiliarity means confirming that all of them have been provided — and their location, since documents may come by email or, where security or the client's guidelines require, have to be retrieved from a shared folder or document management system; confirm at the beginning that every link, access and permission to download is in place, to prevent delays.
- Perform a gap analysis to verify that everything needed is in hand, which a procedure or checklist can guide where the document type is unfamiliar.
- Perform the check, which in a thorough form takes in consistency checks, editing and content verification.
- Document the check as the applicable procedures require, which may mean completing a checklist or signing a review form.
Although the check is often done in a short time because of tight timelines, a few moments spent taking stock at the start of the review can pay dividends. The reviewer establishes how much there is to do and how long there is to do it, and asks the author whether any part has already been through quality control and does not need rechecking — the non-data-driven sections of a clinical study report are often reviewed before the results section is written — and whether all the checklist items still apply. The reviewer also establishes whether the review is being split with another reviewer. The whole checklist should be read before starting, because the time each item takes is easy to misjudge: the easy items, such as the consistency of headers and footers or the use of page breaks, can go quickly and feel like progress, and then one item can require checking the entire document against its source. Reviewing the checklist first gives a better feel for how long the check will take and which items to do first.
Time for the check
The time a check needs is assessed before a deadline is committed to. New editors need extra time while they learn, and even for an experienced editor the time varies with the document type, the complexity of the indication, the scope of the editing tasks and the writer's skills, which are assessed up front. Too little time for the check during development or at the end of a project can carry considerable quality risk, so the start of the project is when the time is estimated or prioritised, from the checks needed, internal or external, the page count and the document type — a 200-page original protocol developed with several reviews involving an external subject-matter expert may need far more adjudication time than an amendment that only clarifies one exclusion criterion — along with the overall timeline, the complexity of the content and the resources available. Tight deadlines are not to be allowed to affect the document's quality: a strategy is mapped out to meet them, and having a plan in place is itself a stress reliever.
Where the time is too short, a second reviewer can be brought in and the work divided. Where timelines are very tight, two reviewers can work in parallel on different parts of the checklist or different parts of the document, and it is better to suggest bringing in another reviewer at the start than to discover the shortfall as the deadline approaches; priority areas can also be agreed in case time runs out. Where the deadline is non-negotiable, the same up-front assessment decides whether a second editor is needed, and if one is, the tasks are divided by logical topics and possibly by the editors' skill sets. Where time is short, the reviewer can also offer to send comments to the author in stages so that revision can begin.
The quality controllers surveyed in 2018 named unreasonable or limited timelines and problems with source documents — inconsistent annotation, and writers not providing the materials or saying where they were — as their top frustrations. Practitioners working in small groups at a 2018 forum proposed solutions for managing timelines and for managing problems with source documents. Where there is time to prepare, they proposed that a clear process be developed in advance and writers and quality controllers trained in it; that writers be trained in annotation style, and quality controllers communicate any changes they make to annotations; that the quality controller be involved from kickoff, so that they can say how long review will take; that metrics be used to build formulas for typical and for more complex project timelines; and that the end client understand and keep to its own timeline for feedback, with who gives feedback to whom, and when, clearly defined. Where there is no time to prepare, they proposed checking the document to predict a deadline that can be promised, splitting complex work with others, making a priority list that separates what must be done from less important issues, establishing what can and cannot be done in the time, sending the document back in chunks so that the writer can address comments while the check continues, and considering telling the client that a larger budget may be needed for more quality controllers to meet an expedited turnaround. For source documents, they proposed that shell folders be created for all source materials at the outset, with each person who passes the document on responsible for the right documents being in the right place and no folder being empty; that every writer, medical editor and quality controller be trained in the annotation style, fact-checking procedures and quality control and editing tasks, and in one consistent way of saying where the latest sources are; and that a master list of resources, perhaps hyperlinked, give quick access to the source documents. The common thread in every solution was communication.
Reading the document and returning comments
PPD sets every new editor two editing rules. The first is that nothing may be changed that cannot be supported with a source. Editors do not change a word because one sounds better than another; every change must be supported with a grammatical or logical explanation, and where wording is unclear the editor rewrites the sentence and asks the writer whether that is what was meant. The editor also edits for consistency, so that the reader stays focused on what the text says rather than wondering whether it means something else because similar terms are used for the same idea, and the training uses specific terms to show what that means: do end of treatment and end of study refer to the same time point?
At PPD the review is done electronically with changes tracked, and a hard copy is kept so that all of the text can be confirmed as verified, experienced editors checking off the sections and data they have reviewed with pens or markers. Every document must be read for content consistency, grammar and style, and one carrying data to verify needs a further level of review: switching from reviewing words to verifying numbers means adjusting one's thinking, because the source documents are different and so is the text the information is presented in. PPD leaves the order to each editor — all the data first, or the data alongside the text section by section — because comfort levels vary, and what is required is a thorough review within a limited time. Before the review is completed, the editor works through a checklist of items that should have been done during the edit, such as whether headers were checked and hyperlinks verified, because there is much to confirm in editing a document and the list is the reminder.
Two cautions apply to the reading. The check typically comes after several rounds of review, so the text may have been written by several authors: unless something is inaccurate or grammatically incorrect, the reviewer follows the style guide and does not inject personal preferences about word placement, style and flow, and where there is no style guide makes the document consistent with itself, in word choice such as subject against patient and in elements such as headers and footers, because the primary job is ensuring accuracy and consistency, not rewriting or restructuring the document to personal preference. And the check requires complex multitasking, so a reviewer doing it for the first few times may find it easier, if more time-consuming, to verify the content first, then read for spelling, grammar and punctuation, and then again for flow and consistency, with the ability to do all of it in one read-through developing over time.
The conditions of the review affect what it catches. The check probably has to fit around the reviewer's other work, so the reviewer puts other priorities aside, finds quiet, switches off email where possible, and takes breaks over a long review, because tiredness can lead to small mistakes being missed. In one editor's view, the best way to error-proof the process is one consistent routine that plays to the reviewer's strengths and moves through a document quickly and consistently, supported by the electronic tools at hand — a customised Word toolbar, keyboard shortcuts, and autocorrect entries that type frequent comments — and by checklists, used to the reviewer's advantage and to keep track of metrics worth recording.
How the comments are returned affects whether they are acted on correctly. The edits should be clear and easy to understand, and a brief explanation of certain edits can be valuable: text to be cut from a summary but kept in the main body is the example, where without a note the author may delete it throughout the document. Stylistic changes can be justified by pointing at the style guide, which shows that the edit is not a personal preference. The review can be a learning opportunity for the author and the reviewer alike, so a small number of explanations alongside the comments can be helpful.
PPD's second rule governs the comments, which are provided to the writer in complete sentences, following AMA style, and without personal pronouns. Banning the pronoun avoids the subtle confrontation of "I" against "you": I think this should be…because… and You wrote this, but it should have been… become This should be…because… and Should this be…? The finished document is the result of collaboration and all parts of the project should be completed through teamwork, so any form of conflict should be avoided.
The check of changes after the client's comments
After a client returns a document with changes and the writer incorporates them, PPD runs what it calls a check-of-changes review. Only the changes are of interest, so the entire document is not reviewed again: a comparison document is generated identifying the differences between the last edited version and the current one, and interpreting it takes enough experience that its points are explained to a novice.
The check has two objects. The first is verifying that the client's comments have been addressed. The second is determining that each occurrence of a term or passage requiring correction has been changed throughout the document, which is done by searching for the key terms that were changed and confirming that the text around them is consistent throughout.
Safety narratives
For the review of safety narratives, one practitioner stresses one instruction throughout: find out what the story is. The person checking a narrative is protecting the story and making sure the data that support it are correct. That person makes sure the story makes sense and always keeps the reader in mind.
In the hurried preparation of regulatory submission documents, patient narratives may not always be top of mind for writers and those who perform quality control, and the focus on time and prioritisation often rests with the body of the clinical study report; one editor who trains new editors to check narratives holds that, although the report's body holds the aggregated data that allow conclusions, the narratives are the heart of the document. The same editor's approach likens quality control of a patient narrative to putting the pieces of a puzzle together: the reviewer reads the narrative several times, focusing on different items with each pass. One of the passes is simply reading the story, because taking the emphasis away from the data gives an opportunity to understand what the patient actually experienced. Medical writers owe the patients that their data, and their stories, are told correctly and reflected accurately.
The check runs against a variety of source documents, templates and guides. They include the narrative template, which gives the structure; a sample narrative, created by the writer and the study team; the listings, which hold all the individual patient data; patient profiles; the CIOMS or MedWatch forms describing each adverse event; correspondence from investigators, which can add to the story; the clinical study report in progress, with which the narratives must be consistent; the protocol, which says what the patients were supposed to experience and is to be read; the safety narrative plan, which says what narratives will be written; the client's style guide, to be asked for if not supplied; and a quality control checklist, which the reviewer is advised to develop for themselves.
Narratives usually arrive in batches of five or ten, and the reviewer should read the whole batch at once and establish what happened, how each patient was treated and how each story ended — "This is not creative writing, but we are telling a story". Consistency is then looked for and created before the data are checked. The consistency pass looks at style and presentation: the organisation of the narrative, chronological or thematic, paragraph length and spacing after full stops, and the use of American or British English, SI or conventional units, the number of decimal places, whether reference ranges and their sources are given, calendar dates or study days, and generic or trade names.
The data pass has its own mechanics. The reviewer compares the narrative and its sources on two monitors and highlights what has been checked, electronically or on paper; one reviewer finds PDF easier to search than Word. The reviewer asks whether the database has been locked, because if it has not a final check may be needed. The listings, quantitative and from the statisticians, and the CIOMS forms, qualitative and from the investigators, should match, and the listings are the primary source for patient identification, MedDRA terms, medication start and stop dates, and relatedness to study drug.