Clinical regulatory documents tend to be long and complex and filled with data. However, as a regulatory reviewer famously said about clinical documentation: “I usually do not see good documents. What we seem to get are pages and pages that are not directly related to the question. It is hard to find the time to go through all of that stuff.” Nevertheless, these documents also need to clearly and succinctly summarize and explain huge amounts of data, and this is perhaps the most difficult task that an author can face. In particular, the difference between simply presenting data (storage) versus using it to make a compelling argument (story), as is necessary for the regulatory authorities, is often not understood or practiced. This is a general problem in clinical writing where an encyclopaedic style of writing seems to be the standard.
Against this long tradition is the fairly recent concept of “Lean Medical Writing” (LMW) which is becoming increasingly widespread throughout the pharmaceutical industry. What exactly is the purpose of LMW and how is it implemented and used? Although it might appear from the name that LMW involves just using fewer words or less data, I would like to propose that rather than just “thin” medical writing (fewer words and lass data), lean involves much more than this. In addition, the nature of regulatory documents dictates that they must provide comprehensive information while also being readable and summarizing a huge amount of information. This leads to the paradox that such documents should be complete whilst at the same time also being readable. The solution to this is LMW which includes good use of document hierarchy and navigational tools.
A large component of this problem is that authors appear to be concentrating solely on making sure that absolutely every detail is included, and often repeated, rather than in providing it in a format that can be read and understood. The results are huge, obtuse, and extremely difficult to read documents that are user-unfriendly in the extreme. In contrast, the aim of lean medical writing is to produce documents that have all the necessary information but in a form that is easy to access, read, and understand.
So first, how can we resolve the paradox and give reviewers access to all of the data collected while also presenting concise and informative summaries. The key to this is the carefully considered use of headings and tables of contents and to avoid repetition by using cross-referencing generously and wisely. When asked what part of a clinical document regulatory reviewers read first, most authors think that it is the tables and figures, summaries, conclusion, or results. But from my interactions with regulatory reviewers (both active and former), the first they look at in a document is the table of contents. This is because they do not read these documents like a detective novel, starting at page 1, but rather they are looking to find certain information. With such long and complex documents, most reviewers want to find where a particular section of interest is – and to do this they need a table of contents with informative and useful headers.
But beyond the structural organization of the document, there are a number of things that we need to consider for the text itself if we want to have a readable document with LMW. To start, we need to avoid a number of poor writing techniques that are guaranteed “killers” of good LMW: repetition, inconsistency, lack of focus, and useless or irrelevant text.
Repetition is to some extent an unavoidable aspect of large, complex clinical documents (repeating parts of the results conclusions text in the synopsis of a Clinical Study Report [CSR], for example). However, many authors seem to feel that repetition is almost a goal in and of itself. One of the most frustrating things I often see is where a small table in the text is used to present some data followed or preceded by all of the same numbers in a text. I was once told by an author that they had learned that the results should be described in the text. I replied that the results should be described, not repeated. Another author told me that repeating the numbers in the text allows both readers who prefer to read numbers in text or tables to each have their preferred form. I wondered whether there were any people in the world who prefer to read numbers in text! Repeating the numbers in the text does not aid comprehension or make the story clearer, it simply makes the document longer and increases the QC time and expense. In this day of reading documents electronically and the use of cross-linking, there is absolutely no reason for repeating things when one can simply insert an electronic cross-link into the text.
However, as much as saving time and money in finalizing a CSR are worthwhile goals in and of themselves, there is an even more important reason to practice LMW and avoid just repeating lots of numbers in the text. I have seen final documents where the authoring team simply repeated all of the numbers of the in-text table in the text and apparently were so intent of doing this, they missed an important message in the data itself. One of the most consequential examples of this was a study that had a number of different treatment groups, of which 2 were compared for the primary objective. The demographic data section had text and an in-text table with most of the data (age, sex, BMI, race/ethnic group, and several other parameters). The text repeated all of the values and duly noted (among many other numbers) that the mean age for one of the 2 groups was 54 years of age and for the other was 67 years of age, along with standard deviations and ranges. There was no mention of the fact that based on the mean age, one group was primarily elderly (age >65 years) while the other was not.
This difference is crucial to interpret the efficacy and safety results. One can easily imagine how this critical difference in age would affect the frequency of adverse events or deaths. Sure enough, there were more events and deaths in the elderly group, but whether they were due to the difference in study treatment or just due to the age difference was impossible to know. The demographic text should have alerted the reader to this difference in age and then referred to this in the safety and discussions sections. Instead, the data were all there, but no real interpretation was made, i.e., there was storage but no story.
I believe this happens for several reasons. First, authoring teams are always under time pressure, and the motivation to just fill the CSR without worrying too much about the meaning can be very strong. Some people are actually worried about interpreting the data incorrectly and thus write defensively rather than to actively communicate, only giving numbers and data so as to avoid making a mistake. Defensive writing is when authors are frightened of drawing any conclusions or saying anything “wrong” about the data and as a result, they don’t end up saying anything. This unfortunate habit is often a holdover from academic research where there is a widespread feeling that the only thing worse than saying nothing was saying something wrong or stupid, and thus many people never said anything. I remember a saying when I was studying that doctoral theses were graded by throwing them down a staircase, with the heaviest (the one with the most pages and thus going farthest down the stairs) getting the best grade. Many organizations continue this anti-intellectual climate which so inhibits intellectual progress. Finally, it can be simple mental laziness, as just cutting and pasting data from statistical output is much easier than thinking about what it might mean.
At the other end of the spectrum are authors who lose sight of the fact that they are writing a scientific-technical document and attempt to make their prose like literature. I even had one member of an authoring team complain about a paragraph as not being elegant. I attempted to explain that the reviewers at the regulatory agencies do not read these documents for pleasure and that attempting to make the text more “interesting”, by, for example, never using a word twice in the same paragraph, leads to inconsistencies that can impede comprehension. This is a case of too much rather than too little (as in the above case), but the result is the same – text that is not fit-for-purpose and misses its target.
Thus, the real goal of LMW is to create regulatory documents that are appropriate for their intended audience. Reviewers are not reading these documents to just get lots of data – they could simply look at the raw statistical outputs if that were the case (where the data are stored). Neither are they reading for enjoyment and thus attempts to make the text more interesting to read are misguided. Clinical documents should be functional, not beautiful. They should not be propaganda, but they should present what the data mean, tell a story, and direct the reader easily to all of the data that support these conclusions. Most importantly, they need to address the question that the study and its design aimed to answer and make it clear what that answer is. This way medical writers can make the jobs of the regulatory authorities easier and allow them to make the decisions necessary to get new life-saving medications to patients and the public as rapidly as possible.