Writing about a real-world study

A real-world study is written up for readers who range from regulators and payers to patients. Writing it well rests on building the narrative for them, having every study detail before drafting, reporting the data, methods, bias and missing data transparently, protecting the people in the data, choosing language that does not overstate causation, and weighing the study's strengths beside its limitations.

Aim and audience

Real-world data can address a broad range of questions, so a variety of stakeholders are interested in the answers, and it helps to consider the needs and expectations of the audience. ¹ The content a medical writer develops includes study reports, integrated evidence plans, reimbursement submissions, regulatory documents, internal training materials, publications, information for patients and medical communication materials. ¹ The format and the audience influence how the evidence is presented, and the writer should also consider how stakeholders will use the content in the future. ¹ Audiences differ in expertise, and some may not be medically trained, so communications need to be tailored to them. ¹ The language may differ with the audience's expertise, but the writer's role is to tell a compelling story that resonates with the reader, whether pharmaceutical industry professionals, regulators, payers, healthcare professionals or patients. ¹ It is important to think carefully about how to build a narrative that shows the audience how the answer to a real-world question is relevant to them and why the evidence matters. ¹

Two questions to ask when writing about a real-world study concern its aim and its audience. ²

• What is the aim of the study? Any communication needs a compelling story. Think about why the study was done and the real-world question it is trying to answer, and about what background the reader needs to appreciate its relevance and importance.
• Who is the intended audience? Audiences differ in expertise and priorities. Avoid jargon, explain concepts in simple, unambiguous terms, and have the appropriate template for documents such as trial protocols and clinical study reports.

^{2, 1}

Study details and data sources

As well as being aware of evolving guidance, it is increasingly important for the writer to understand how to work with real-world data and the unique insights they can provide. ³ Writing about a study accurately and highlighting its strengths and limitations requires understanding how it was conducted and why its methods were chosen, and transparent explanation of how the data sources were selected and the study designed is essential. ⁴ Before writing starts, it is important to ensure that the source materials provided contain all the necessary study details. ⁴ Reporting guidelines for non-interventional studies, such as STROBE, which has separate checklists for cross-sectional, case-control and cohort studies, are useful for checking what a manuscript needs to include, HARPER supports the transparent reporting and reproducibility of real-world study protocols, and any guidance that applies should be read before work starts. ⁴ Recognised reporting standards such as STROBE and RECORD should be considered, as they can help to enhance reproducibility and regulatory acceptance. ⁵

Four questions to ask concern the source information. ⁶

• Are all the study details and data to hand? Check guidelines such as STROBE and HARPER for what a real-world study should report. The field is evolving rapidly, so watch for new guidance and templates that support data transparency. Ask the data owner as soon as possible for anything missing or unclear.
• What data sources and endpoints were used? The sources are numerous and can be very different. Understand them and how they are formatted well enough to explain them, be clear how the endpoints relate to the study question, and give enough context in the methods and results for the reader to assess the data and understand their meaning.
• Do any details need removing to maintain patient anonymity? The data reflect real people. Mask or remove any detail that could compromise anonymity. Sources such as interviews and free text in questionnaires may contain information that needs reporting sensitively.
• Were patient consent and ethics approval obtained? Ethics processes for real-world studies may be less straightforward than for clinical trials. Include some statement on ethical review, and check the requirements with the data owner where they are unclear. Consent and approval may not apply where fully anonymised data were used, and the exemption should then be explained clearly.

^{6, 4}

Patient anonymity and ethics approval

Real-world sources often hold patient data that were not collected with research in mind and may not be fully anonymised, so care must be taken to report them ethically. ⁴ Manuscripts should confirm informed consent where appropriate and give either details of ethics committee approval or an explanation of why none was required. ⁴

The purpose of transparent reporting

Transparency is vital for the credibility, reproducibility and ultimately the regulatory acceptance of analyses based on real-world data. ⁷ Transparent reporting, including clear documentation of study populations, data sources, statistical methods and limitations, is critical to maintaining scientific rigour, particularly when seeking regulatory acceptance. ⁵ Enhancing transparency strengthens confidence in real-world evidence and facilitates its successful integration into clinical decision-making and regulatory assessments. ^{8, 9} Enhanced transparency builds confidence among regulators and the scientific community and paves the way for the broader integration of real-world data in healthcare. ⁹ By reporting real-world studies effectively and transparently, medical writers can support the pharmaceutical industry in building trust in the diverse and valuable insights gained from real-world data. ¹⁰

What transparent reporting covers

Several reporting strategies are important for transparency when summarising evidence from real-world data, each for the reason given. ^{7, 11}

• Data sources and their quality. Give the origin in sufficient, clear detail, naming the specific sources, such as electronic health records, claims or registries, and saying whether the data were collected for a specified purpose or extracted from a database without one. Describe how the data were extracted or collected, managed, cleaned and processed, with the steps taken to maintain integrity and quality. Details of collection and curation can give important insights into limitations or potential biases inherent to the data, and the reader should be sufficiently informed before making decisions or interpretations based on the results.
• The patient population. Give the inclusion and exclusion criteria, with attention to specific characteristics. For registries, give the selection criteria for the whole registry and separately for the subset analysed. Where real-world data supplement a trial, detail the differences between the populations. Distinguish clearly between mining a whole registry and selecting from it on prespecified criteria. Understanding the source population helps reviewers understand how generalisable the findings are and how representative they are of any targeted populations, and without randomisation, treated patients may differ systematically from untreated ones.
• The analysis plan and methods. Describe the statistical methods fully, with every assumption stated, justify the choices and any alternatives considered, and report the sensitivity analyses run to assess the robustness of the findings. Transparency about the methods ensures reproducibility, a basic tenet of rigorous research, and regulators require transparency in how real-world studies are conducted.
• Bias and confounding. Name the potential sources, such as selection bias and confounding, say how adjustments minimised them, and present results before and after adjustment to show their impact. Real-world data come from routine clinical practice, where factors such as disease history, prior treatment and clinical setting can influence outcomes and introduce bias when interpreting treatment effectiveness.
• Endpoint definitions. Standardise them across sources, sites and institutions, and give the criteria for derivation and the validation process. Sources may define an endpoint differently, which can translate into a different outcome or endpoint in the analysis, and the degree of consistency is essential for interpreting the results.

¹¹

Data collection outside a randomised controlled trial is inherently variable, so endpoints in real-world studies may be more complex than in trials, particularly where data were collected over long periods during which diagnostic codes, assessment methods, guidelines or policies changed. ⁴ Where data on a specific clinical variable are unavailable, surrogate or composite measures may be used to determine key outcomes indirectly. ⁴ Subgroup analyses and missing data are common, so reported outcomes need a clear follow-up timeframe and a clear account of the patients in each analysis. ⁴ The writer needs to think about the nature of the data being reported, how each variable relates to the question and what the reader needs to know to understand the results. ⁴

Transparent reporting should also cover missing data: the extent of missingness and its potential impact on interpretation, the handling method, such as complete-case analysis or the type of imputation, and sensitivity analyses of the missing-data assumptions. ⁵ The strategies used to minimise bias and handle missing data should be described in the methods. ¹²

Wherever possible, the protocol and analytical code should be made accessible, as this greatly enhances reproducibility, and open-source platforms are a good place to share them provided they safeguard patient privacy. ⁵

Causal language

Real-world studies have greater potential for bias than randomised controlled trials, so clear reporting of the statistical methods is paramount for building trust that the conclusions are robust. ^{12, 13} As for any study, objective reporting and interpretation of the findings is essential. ⁸ Non-interventional studies are often designed to find associations rather than to conclude causality, so causal language needs caution, particularly when findings rest solely on descriptive statistics. ⁶ Given the inherent difficulty of establishing causal inference from real-world data, careful reporting of the analyses matters, and taking care not to overstate causal relationships is essential, especially given the observational nature of the studies. ⁸ Communicating the findings requires measured and balanced language, with ample discussion of potential biases and acknowledgement of any limitations, and remaining uncertainties should be highlighted and, where meaningful, alternative explanations presented. ⁸

Strengths, limitations and generalisability

Real-world evidence is often caveated with a list of issues, that must be addressed, such as data quality, bias and a general lack of methodological rigour, which make it sound challenging, but it is worth the effort. ¹⁰ In a manuscript's discussion, being clear about generalisability, for example that a study looked only at patients of a certain age or ethnic background, does not diminish the study's validity but gives the reader essential context. ⁶ Discussing limitations is always important, and the inherent limitations of real-world studies should be acknowledged. ⁶ Specific limitations of design, data integrity or interpretation should be discussed with the authors to agree how to address them and whether further studies are warranted. ⁶ Real-world data are by nature variable and complex, but real-world studies give insights that trials cannot, and any limitations should be considered in the context of the study's strengths. ⁶

Five questions to ask concern the interpretation of the data. ¹³

• What steps were taken to minimise bias? Check the protocol and statistical analysis plan if available, and ask the data owner if unsure. Include any sensitivity analyses conducted and the results of other analyses that support the robustness of the data.
• Have the outcomes been described appropriately? Choice of language is important. Real-world studies show effectiveness, not efficacy. Avoid language of causality or association where findings rest solely on descriptive statistics.
• What are the study's strengths? Non-interventional studies examine what happens in real life in a way that clinical trials cannot. Highlight the strengths, not just the limitations, and ask the authors if unsure what they are.
• What are the study's limitations? Discuss them transparently, considering potential sources of bias and the inherent limitations of real-world data, and agree with the authors how to address them.
• How generalisable are the findings? The reader needs to understand what the findings mean in a broader context. A statement on generalisability is always important to include. Think about how representative the patients' demographics and clinical characteristics are of the global patient population.

¹³