Clinical trial design: Considerations for medical writers developing protocols

Tiago Silva have been made to harmonise the structure of not available. This creates an additional challenge IQVIA, Porto Salvo, Portugal trial protocols (e.g., TransCelerate Common for writers. Template 2 and SPIRIT Statement 3), a This article reviews the main concepts standard similar to ICH E3 4 for trial reports, affecting trial designs presented in proto cols and Correspondence to: defining what information to present and how is how to address them from a medical writing Tiago Silva Sr Medical Writer Global Medical Writing Table 1. Protocol writing tips and considerations for key sections IQVIA Lagoas Park Introduction Edifício 3 – Piso 3 l The introduction section should include a literature review covering the indication, 2740-266 Porto Salvo available therapeutic options, and test treatment(s) Portugal l Present the scientific rationale of the trial, identifying its primary purpose, what it aims to +351 219016549 achieve, and its importance [email protected] Objectives l Confirm consistency between the rationale and objectives Abstract l Write SMART objectives (see Table 2 opposite) Clinical trial protocols must provide a clear trial design to meet the study objectives. Population Medical writers must understand and be l Define the target population and present a list of inclusion/exclusion criteria ready to review and discuss aspects of the trial l Confirm that the inclusion/exclusion criteria are consistent with the objectives design with the protocol development team, in order to write a clear and accurate protocol. Endpoints This article reviews some of the main trial l Write short descriptions of the endpoints, which should be measurable design concepts medical writers should l Confirm consistency between the objectives and endpoints expect to find when writing protocols. l Confirm that all variables are captured in the schedule of assessments l Confirm consistency between the defined estimands, objectives, endpoints, and analyses

Introduction Trial design Clinical trial protocols for trials evaluating l Confirm that the trial design is clear and consistent with the objectives, endpoints, and schedule pharmacological products are complex docu - of assessments ments that describe the medical, ethical, and l Confirm that bias minimisation methods are presented. If randomisation and/or blinding are regulatory foundations of the trial. Medical not used in a comparative trial, this should be justified writers work together with protocol development teams of subject matter experts (including Control groups medical experts, statisticians, regulatory experts, l Confirm that the control group is clearly identified and justified operational experts, and pharmacokineticists) to l Confirm that the control group is aligned with the design and objectives write clear protocols that will address the l Confirm that all test treatments (including placebo) are clearly identified and characterised proposed medical questions and protect partici - pant safety and rights. To accomplish this, writers Statistical considerations must understand and be able to communicate l Confirm consistency between statistical considerations and trial endpoints clinical trial design concepts that are often l Confirm that all variables analysed are collected at the appropriate times in the schedule of complex. Moreover, although International assessments Council for Harmonisation (ICH) Good Clinical l Avoid too much detail when presenting the statistical analysis methods, referring to the Practice (GCP) 1 provides recommendations on statistical analysis plan when appropriate what a trial protocol should include, and efforts

56 | March 2020 Medical Writing | Volume 29 Number 1 Silva – Clinical trial design

perspec tive while ensuring compliance with ICH Objectives in clinical trials can be divided into Population GCP. Operational, regulatory, and ethical three categories: 7 Protocols should briefly define the target concepts are not discussed. l Primary (typically one): aims to directly population (i.e., the set of people throughout the Table 1 summarises some writing tips for each answer the primary purpose of the clinical world for which the trial results may be key section of the protocol. trial generalised). 8 A detailed list of inclusion and l Secondary : other actions relevant to and/or exclusion criteria should follow, specifying: 8,9 Rationale indirectly associated with the rationale l Demographic characteristics (e.g., age, sex, Clinical trials assess the efficacy, safety, and/or l Exploratory : hypothesis-generating objec- body mass index) pharmacological characteristics of medicinal tives that can be confirmed in dedicated studies l The medical indication under study, products in human participants. The protocol must present a rationale that identifies the primary purpose of the trial, usually in the Table 2. Example of a SMART written objective “Introduction” section (Table 1). At trial con - ception, the protocol development team should To assess the efficacy of paracetamol 1000 mg per os versus placebo in adult patients with fever of consider: unknown origin (body temperature >37.5°C) 2 hours after administration. l If the rationale is clear l If the trial is clinically relevant and feasible Specific: identifies the action (to assess efficacy), medication(s) (paracetamol and placebo), l If an unnecessary risk/burden will be posed population (adults), and indication (fever of unknown origin) to trial participants Measurable: temperature is a measurable variable

Objectives Achievable: feasible and not burdensome to the participant (to be confirmed by the trial medical Trial objectives are the actions proposed to fulfil expert) the trial’s rationale. The objectives should be Relevant: objectives should be clinically relevant (to be confirmed by the trial medical expert) conceived by the protocol development team and written by the medical writer according to the Time based: in this example, a 2 hour timeframe is specified SMART principles (Table 2). 5,6

www.emwa.org Volume 29 Number 1 | Medical Writing March 2020 | 57 Clinical trial design – Silva

acceptable/prohibited comorbidities, and l Mean body temperature 2 hours after test acceptable/prohibited concomitant medi - treatment administration cations l Ethical requirements for participation (e.g., The protocol development team is responsi ble informed consent) for defining the endpoints, while the medical l Exclusion criteria that may bias result writer should ensure that they are clearly written interpretation or pose an unnecessary risk to (Table 1). the participant Endpoints are classified in accordance with the corresponding objectives (primary, sec - Endpoints ondary, exploratory). The primary endpoint In clinical trials, variables are the parameters should provide the most clinically relevant and (sociodemographic, clinical efficacy and safety, convincing evidence directly related to the laboratory/imaging-related, etc.) that will be primary objective of the trial and its selection measured. Variables can be independent (vari - should reflect the accepted norms and standards ables that potentially influence health outcomes, in the relevant field of research. 11 e.g., treatment, age, sex, smoking history) or Endpoints should be measurable and should dependent (health outcomes, e.g., vital signs, be concisely described in the protocol, while laboratory parameters). 8 details of how the endpoints will be An endpoint is defined by The captured can be provided in one or Spilker as “ an indicator meas ured protocol defines more dedicated protocol l Population-level summary statis tics for in a patient or biological sample the procedures by which sections. All endpoints and the endpoint: the basis for treatment to assess safety, efficacy or a clinical trial is conducted, variables must also be comparisons another trial objective .10 Each captured in a schedule of A summary of the use of estimands in clinical objective must be matched and its writing requires careful assessments that identifies trials can be found in a publication by Bridge and with at least one endpoint, assessment of all concepts which variables will be Schindler. 13 to ensure that the right surrounding trial design for measured at each trial visit variables will be captured and accuracy and and that provides sum marised Trial design that all questions posed by the procedural information for The trial design will dictate participant treatment trial will be addressed. As an consistency. quick reference to the trial site and follow-up, the number of treatment groups, example, let’s consider the following teams. and data collection, among other aspects. objective: Estimands were recently introduced in l To assess the efficacy of paracetamol 1,000 mg ICH E9 (R1) and describe the treatment effect Single group design per os versus placebo in adult patients with fever with consideration of specified post- In single group trials, variables are compared in of unknown origin (body temperature >37.5°C) randomisation events and whether the outcome the same participant before and a certain time 2 hours after administration of the test treatment would be under different conditions. 12 Four after exposure to the test treatment (intra - interrelated attributes are considered for this participant analysis). 8 These are typically early Two possibilities for the corresponding endpoint purpose: 12,13 phase trials or are conducted where limited can be proposed (the best possibility should be l Population: participants targeted by the trial participant pools are available. Results are usually selected with the protocol develop ment team l Variable/endpoint preliminary since it is not possible to blind the based on clinical relevance and feasibility): l Post-randomisation events: events that treatment. 14 l Proportion of patients with a body temperature happen to participants that may affect results higher than 37.5°C 2 hours after test treatment (e.g., death, treatment discontinuation, use of Comparative design administration rescue medications) In comparative trials, two or more test treatments

s Washout Washout Comparator Comparator Variable

Group A t Drug A Comparator Sample l l l

Group B t Drug A Washout Drug A

Time t Figure 1. Crossover design Figure 2. Ideal crossover trial variable

58 | March 2020 Medical Writing | Volume 29 Number 1 Silva – Clinical trial design

Table 3. Factorial design (2 2) regarding key baseline variables. 16,17 After ensuring that all treatment groups are Product A Product B comparable, it is important to confirm that Yes (+) No (–) participant follow-up and outcome assessments Yes (+) Group 1 ++ Group 3 + – are not biased. 8 The key trial procedure here is No (–) Group 2 –+ Group 4 –– blinding (or masking): the process of ensuring that the Investigator and/or participant is and complementary effects. However, they can unaware of the treatment assigned. The main be complex if a high number of groups is types of blinding include: 15 required. 14 l Double blinding: both the Investigator and Minimising bias in comparative trials the participant are unaware of the treatment Comparative trials typically include measures to given. This is the preferred type of blinding as minimise bias and ensure groups are comparable. it avoids biased assessment of outcomes by Randomisation , in which trial participants are both the Investigator and the participant randomly assigned to the different study l Single blinding: only the Investigator or the treatments, is the standard method to obtain participant is aware of the treatment given treatment groups with similar baseline charac - l Open label: both the Investigator and the teristics (e.g., age, sex). Randomisation increases participant are aware of the treatment given the likelihood that baseline characteristics that are compared. In a parallel design , participants could confound treatment effects will be Double blinding requires test treatments to be will keep the same treatment until trial distributed equally among treatment groups. 8 indistinguishable (shape, colour, smell, taste), completion or discontinuation. At defined time Randomisation lists with the participant codes which is sometimes not possible (e.g., comparing points, variables are compared between groups and respective treatment allocations are infor - two active treatments with different formu - (interparticipant analysis). 15 mat ically generated by different methods: lations). One way to overcome this is to perform In a crossover design , participants receive all l Simple randomisation: based on a single a double dummy trial , where each treatment has treatments in a randomised sequence. Adminis - sequence of random assignments (like tossing a matched placebo (dummy), so each participant tration of different treatments is separated by a a coin), it is useful for large samples, but can receives one active treatment and the placebo washout period, where no active treatment is create unequal treatment groups in smaller version of the comparator. 15 given. This is done to prevent the effect of the samples. 15-17 previous treatment from biasing the new l Block randomisation: blocks of equal size are Control groups treatment, i.e., carryover effects (Figure 1). 8,15 defined with all possible treatment orders and The choice of the control group in comparative Because the same participant receives different are picked randomly to generate the random - trials should consider the rationale and test treatments, intraparticipant analyses are isation list (Figure 3). 8,16,17 objectives, along with regulatory, operational, possible. l Stratified randomisation: the sample is and ethical aspects. Crossover trials are useful when assessing stratified by key baseline characteristics (e.g., stable variables and conditions that produce a sex, age). Participants are randomised so that Placebo response during treatment and return to near- these characteristics are distributed equally Placebos are formulations without any active baseline levels during washout (Figure 2). 8,14 between treatment groups. 8,16,17 pharmaceutical ingredient used in double blind trials. 15 These must be indistinguishable from the Factorial designs compare different treatments Another allocation method is adaptive random - test treatment in terms of packaging, labelling, given as monotherapy or in combination, isation, where the first participant is randomised, size, shape, opacity, coatings, viscosity, colour, creating groups for each possibility (Table 3). while subsequent participants are allocated non- smell, flavour, and route of administration. 18 These trials assess various possible interactions randomly to minimise group imbalances Placebos are useful to minimise bias and distinguish the real effects of the test treatment and “noise” effects, such as: 15 AABB ABAB ABBA BABA BBAA BAAB l Normal physiology and spontaneous fluctuations l External factors that can alter the participant

t t t response (e.g., increases in liver transaminase A B A B B B A A A A B B Participant no. 1 2 3 4 5 6 7 8 9 10 11 12 levels after a long period of hospitalisation in a clinical trial unit due to a strict diet, lack of Figure 3. Block randomisation for treatments A and B exercise, or other lifestyle changes) Twelve participants are randomly allocated to treatment A or treatment B using three randomly On the other hand, placebo groups may not be selected blocks of 4, each containing a unique sequence of treatment options. appropriate in some settings, such as serious

www.emwa.org Volume 29 Number 1 | Medical Writing March 2020 | 59 Clinical trial design – Silva

} Inferior } Inferior

Non-inferior } Superior } } Superior } Inconclusive } Inconclusive t t 0 -∆ 0 Effect size (test product – comparator) Effect size (test product – comparator)

Figure 4. Possible outcomes of a superiority trial Figure 5. Possible outcomes of a non-inferiority trial Data are shown as the point estimate and 95% confidence interval. Data are shown as the point estimate and 95% confidence interval.

with available therapeutic alternatives (Figure 5). 19 If the 95% CI lies entirely above 0, l Standard of care (treatment as usual, routine (e.g., oncology, infectious diseases). 15 there is evidence of superiority at the two-sided care): this can be employed to compare the 5% significance level (p<0.05). test treatment with existing practice. Active comparator A justification for the non-inferiority margin, However, “standard of care” can vary between When comparing a test treatment against an typically provided by the trial statistician, should study sites and countries, hampering an active comparator, different trials can be be included in the protocol. objective definition 25 considered. Bioequivalence trials compare two products l Active placebo : a placebo that mimics the Superiority trials aim to demonstrate the with the same active pharmaceutical ingredient adverse effects of the test treatment. This can superiority of the test treatment against the active (i.e., different formulations of the same product be useful when the risk of unblinding due to comparator regarding the primary endpoint. If or generic versus comparator). 22 Two products characteristic adverse events is high 26 the difference between the test treatment and are considered bioequivalent if they produce the comparator and 95% confidence interval (CI) are same plasma concentration-time profiles, i.e., if Statistical considerations higher than 0 for the primary endpoint, the test the 90% CIs for the geometric mean ratios for Statistician support is needed when developing treatment is considered superior (Figure 4). 19 the area under the curve from time 0 to last the statistical sections of the protocol. When Sometimes, a different test treatment may not measurable concentration (AUC 0-t ) and applicable, other experts should also be consulted be superior in terms of efficacy but may offer maximum plasma concentration (C max ) for the (e.g., a pharmacokinetics/pharmacodynamics other advantages (e.g., a better toxicity profile, test and reference products lie entirely within the expert, quality of life expert). The protocol more convenient administration). 15 Here, non - interval 80% to 125%. 22,23 Examples are shown should present a rationale for the calculated inferiority trials can be considered. These trials in Figure 6. sample size with the necessary statistical and evaluate whether the test treatment is as good as clinical assumptions (based on the primary or worse to an acceptable degree compared to the Other control groups endpoint). In addition, the statistical analysis reference. Here, a non-inferiority margin (- Δ in Other potential control groups include: methods should be summarised for all Figure 5) is defined. 8,20,21 If the difference l No treatment : this method is not blindable endpoints, with full details being provided in a between the test treatment and comparator and and can be unethical unless it is confirmed separate statistical analysis plan. 95% CI for the primary endpoint is higher than that participants will not be subjected to an -Δ, the test treatment is considered non-inferior unacceptable risk 24

} Not bioequivalent

} Bioquivalent

} Not bioequivalent

} Inconclusive t 0.80 1 1.25

Geormetric mean ratio (AUC 0-t and C max ) (test product – comparator)

Figure 6. Possible outcomes of a bioequivalence trial Data are shown as the geometric mean ratio and 90% confidence interval for the area under the curve from time 0 to last measurable concentration (AUC 0-t ) and maximum plasma concentration (C max ).

60 | March 2020 Medical Writing | Volume 29 Number 1 Silva – Clinical trial design

Final remarks Clinicians, Educators, and Researchers. 19. Food and Drug Administration [Internet]. The protocol defines the procedures by which a 2nd ed. New York: Springer New York; Non Inferiority Clinical Trials to Establish clinical trial is conducted, and its writing requires 2011. Effectiveness – Guidance for Industry. 2016 an assessment of all concepts surrounding trial 7. Bacchieri A, Cioppa GD. Fundamentals Nov [cited 2020 Jan 05]. Available from: design. Medical writers must understand these of : Bridging Medicine, https://www.fda.gov/media/78504/ concepts and work with the protocol develop - Statistics and Operations. Milan: Springer; download. ment team to communicate them clearly. This 2007. 20. Schumi J, Wittes JT. Through the looking will protect the scientific integrity of the trial and 8. Hulley SB, Cummings SR, Browner WS, glass: understanding non-inferiority. Trials. the safety of the participants. Grady D, Newman TB. Designing clinical 2011;12:106. research. 4th ed. Philadelphia: Wolters 21. Althunian TA, de Boer A, Klungel OH, Acknowledgements Kluwer; 2013. Insani WN, Groenwold RHH. Methods of The author would like to thank Deborah 9. Patino CM, Ferreira JC. Inclusion and defining the non-inferiority margin in McPhail, Isy Goodwin, and Robert Schall for exclusion criteria in research studies: randomized, double-blind controlled trials: their comments and insight. definitions and why they matter. J Bras a . Trials. 2017;18(1):107. Pneumol. 2018;44(2):84. 22. Atkinson AJ, Huang SM, Lertora JJL, Disclaimers 10. Spilker B. Guide to Clinical Trials. Markey SP. Principles of Clinical The opinions expressed in this article are the Philadelphia: Lippincott-Raven Press; 1991. Pharmacology. Amsterdam/Boston: author’s own and are not necessarily shared by 11. International Council for Harmonisation of Academic Press; 2012. his employer or EMWA. Technical Requirements for 23. European Medicines Agency [Internet]. Pharmaceuticals for Human Use [Internet]. Guideline on Investigation of Bio - Conflicts of interest Statistical Principles for Clinical Trials. E9. equivalence. CPMP/EWP/QWP/1401/ The author declares no conflicts of interest. Current Step 4 version. 1998 Feb 5 [cited 98 Rev. 1/ Corr. 2010 Jan 20 [cited 2020 2020 Jan 05]. Available from: Jan 05]. Available from: https://www.ema. References https://database.ich.org/sites/default/ europa.eu/en/ documents/scientific- 1. International Council for Harmonisation of files/E9 _Guideline.pdf. guideline/guideline-investigation- Technical Requirements for 12. International Council for Harmonisation of bioequivalence-rev1 _en.pdf. Pharmaceuticals for Human Use [Internet]. Technical Requirements for 24. Nardini C. The ethics of clinical trials. Integrated Addendum to ICH E6(R1): Pharmaceuticals for Human Use [Internet]. Ecancermedicalscience. 2014;8:387. Guideline for Good Clinical Practice Estimands and sensitivity analysis in 25. Ayling K, Brierley S, Johnson B, Heller S, E6(R2). Current Step 4 version. 2016 clinical trials. E9 (R1). Current Step 2 Eiser C. How standard is standard care? Nov 9 [cited 2020 Jan 08]. Available from: version. 2017 Jun 16 [cited 2020 Jan 05]. Exploring control group outcomes in https://database.ich.org/sites/default/files Available from: https://database.ich.org/ behaviour change interventions for young /E6 _R2 _Addendum.pdf. sites/default/files/E9-R1 _ Step4 _ people with type 1 diabetes. Psychol 2. TransCelerate Biopharma Inc. Clinical Guideline _2019 _1203.pdf. Health. 2015;30(1):85–103. Content & Reuse. 2017 [cited 2020 Jan 08]. 13. Bridge H, Schindler T. Estimands – closing 26. Jensen JS, Bielefeldt AØ, Hróbjartsson A. Available from: the gap between study design and analysis. Active placebo control groups of pharma - https://transceleratebiopharmainc.com/ Med Writ. 2018;27(4):52–6. cological interventions were rarely used but initiatives/clinical-content-reuse/. 14. Evans SR. Clinical trial structures. J Exp merited serious consideration: a methodo - 3. Chan A, Tetzlaff JM, Altman DG, et al. Stroke Transl Med. 2010;3(1):8-18. logical overview. J Clin Epidemiol. SPIRIT 2013 Statement: Defining 15. Griffin JP, Posner J, Barker GR. The 2017;87:35–46. Standard Protocol Items for Clinical Trials. Textbook of Pharmaceutical Medicine. Ann Intern Med. 2013;158(3):200–7. 7th ed Chichester: Wiley-Blackwell/BMJ 4. International Council for Harmonisation of Books; 2013. Technical Requirements for 16. Suresh K. An overview of randomization Pharmaceuticals for Human Use [Internet]. techniques: An unbiased assessment of Structure and Content of Clinical Study outcome in clinical research. J Hum Reprod Author information Reports. Current Step 4 version. 1995 Nov Sci. 2011;4(1):8–11. Tiago Silva has over 8 years of medical 30 [cited 2020 Jan 08]. Available from: 17. Lim CY, In J. Randomization in clinical writing experience in a CRO setting. His https://database.ich.org/sites/default/ studies. Korean J Anesthesiol. 2019;72(3): current role involves the development of files/E3 _Guideline.pdf. 221–32. clinical research documents, such as protocols 5. Al-Jundi A, Sakka S. Protocol Writing in 18. Brody T. Clinical Trials: Study Design, and clinical study reports, across all stages Clinical Research. J Clin Diagn Res. Endpoints and Biomarkers, Drug Safety, of development for several therapeutic 2016;10(11):ZE10 –3. and FDA and ICH Guidelines. Amsterdam: indications. 6. Taylor R. Medical Writing: A Guide for Academic Press; 2011.

www.emwa.org Volume 29 Number 1 | Medical Writing March 2020 | 61