Diabetes Ther (2021) 12:2635–2644 https://doi.org/10.1007/s13300-021-01143-6 COMMENTARY Randomised Controlled Trials in Diabetes Research: A Pathway to Interpreting Published Results Tadej Battelino . Didac Mauricio Received: May 14, 2021 / Accepted: August 9, 2021 / Published online: August 29, 2021 Ó The Author(s) 2021 Keywords: Randomised controlled trial; Publications; Interpretation; CONSORT Key Summary Points Randomised controlled trials (RCTs) remain the gold standard for direct treatment comparisons. However, interpreting the results of RCTs and making judgements about the quality of evidence and how results may be applicable to diabetes management can be difficult for healthcare practitioners (HCPs). In this article, a checklist of the points that we consider most important when reading and interpreting RCT publications is summarised, and may serve as useful guidance for HCPs. T. Battelino University Medical Centre Ljubljana, Ljubljana, Slovenia INTRODUCTION T. Battelino Faculty of Medicine, University of Ljubljana, Although the importance of real-world evidence Ljubljana, Slovenia should not be underestimated, randomised D. Mauricio (&) controlled trials (RCTs) remain the gold stan- Hospital de la Santa Creu i Sant Pau, CIBERDEM, dard for direct treatment comparisons. Over the Barcelona, Spain e-mail: didacmauricio@gmail.com last two decades, numerous publications have presented results of RCTs investigating the D. Mauricio safety and efficacy of therapies for the treatment Faculty of Medicine, University of Vic—Central of type 2 diabetes, for example long-acting basal University of Catalonia, Vic, Spain 2636 Diabetes Ther (2021) 12:2635–2644 insulin analogues [1–14]. RCTs such as these use The research questions being posed in the a robust design and methodology that aims to introduction of an article are also useful to help avoid confounders and determine differences clarify whether the RCT has employed the between basal insulins. However, interpreting appropriate study design and, if not, the impact the results of RCTs and making judgements this may have on the reliability of the results. A about the quality of evidence and how results competent RCT research question will encom- may be applicable to diabetes management can pass the participant population, the ther- be difficult for healthcare practitioners (HCPs) apy/clinical intervention, outcomes, and it because of differing trial methodologies, analy- typically questions the effectiveness of one sis plans, and endpoints. therapy versus another (or placebo) [17, 18]. There are guidelines for those developing publications that report the results of RCTs, Methods Section such as the Consolidated Standards of Report- ing Trials (CONSORT) guidelines [15], and Is the Randomisation Method Reported readers are advised to familiarise themselves and Appropriate? with these if possible. However, we have Randomisation in RCTs is critical for ensuring developed a checklist of points we consider that any known or unknown confounding fac- most important when reading and interpreting tors present in the population being investi- RCT publications (Fig. 1). This checklist is based gated are equally distributed, allowing the on CONSORT, other published literature, and effects of an outcome to be reliably attributed to our own experience, and may serve as useful the treatment [17]. As such, details of the ran- guidance for HCPs. This article will discuss these domisation procedure should be clearly repor- key points in more detail, progressing through ted in the methods section of the publication. the sections of a typical RCT publication. Most trials will use equal randomisation, in This article is based on previously conducted which study groups are randomised in a 1:1 studies and does not contain any new studies ratio; however, trial participants may be ran- with human participants or animals performed domised in a 2:1 or 3:1 ratio for reasons by any of the authors. including cost, ethics, increased statistical power, and the potential to gain extra infor- CHECKLIST ITEMS mation on the treatment [15, 19]. It is therefore good practice to provide the specific randomi- Introduction Section sation ratio. Additionally, the specific method by which participants are randomised to a given group is important when determining the reli- Does the Publication Provide a Clear Rationale ability of the randomisation outcome and the for the Study and Give Specific Objectives extent to which bias may have been uninten- or Hypotheses? tionally incorporated [15]. A technique that When undertaking research involving human provides true randomisation (such as a com- subjects, it is essential that there is a clear and puterised sequence generator) is recommended, justified rationale based on a comprehensive as treatment allocation cannot then be pre- understanding of the scientific literature to dicted, whereas the use of information such as avoid unnecessarily exposing people to the risks date of birth to approximate randomisation associated with research [15, 16]. The intro- may lead to selection bias and is therefore not duction to an RCT publication should highlight acceptable [15, 17]. the objectives of the trial, the gaps in the liter- ature that the study would address, its impor- tance to the field, and it should allow the Were Participants and Physicians ‘Blinded’ authors to present the hypotheses that are being to the Treatment Being Received/Given? tested. There are a number of levels in an RCT at which knowledge of the trial and treatments being Diabetes Ther (2021) 12:2635–2644 2637 INTRODUCTION Does the publicaon provide a clear raonale for the study and give specific objecves or Yes No hypotheses? Clinical studies must be jusfied by a clear raonale, to avoid unnecessarily exposing people to the potenal risks of research, and such jusficaon usually involves a summary of the current literature – subsequently, the overall objecve of the trial, or even specific hypotheses, should be presented. □ □ METHODOLOGY Is the randomisaon method reported and appropriate? Yes No The randomisaon method should be reported in the methods secon of the publicaon – a technique that provides true randomisaon (such as a computerised sequence generator) should be used, whereas use of informaon such as date of birth to approximate randomisaon is not acceptable as it may lead to selecon bias. □ □ Were paents and physicians ‘blinded’ to the treatment being received/given? Ideally, neither the paent nor the treang physician would know which treatment was being received/given, to avoid bias – this is a ‘double-blind’ study design. □ □ Was the study conducted in line with the study protocol? Most trials will have a detailed protocol that specified how the trial will be conducted. Deviaons from the planned protocol can affect the interpretaon or relevance of a study and should be clearly stated in the methods secon. □ □ Are all primary and secondary endpoints clearly defined? Primary and secondary endpoints, as well as the stascal methodology used to analyse them, should be pre- defined as per the trial protocol. This informaon should be clearly presented in the methods, and any analyses that were not pre-defined should be clearly stated as ‘exploratory’ or ‘post-hoc’. □ □ Is the sample size appropriate for the analyses being done? Sample size is usually calculated based on the ability to show a pre-defined and clinically relevant difference for the primary endpoint – the assumpons for the sample size calculaon should be provided in the methods. □ □ RESULTS Are all paents accounted for in the trial? Yes No A trial should clearly show what happened to each parcipant – for example, who received treatment, who was included in the analyses (parcipants should be analysed by the groups to which they were randomised, an ‘intent-to-treat’ analysis), and who disconnued treatment. Ideally, this informaon should be captured in a flow diagram for clarity, and disconnuaons and exclusions should be minimal. □ □ Were the groups similar at the start of the trial? The more similar the comparator groups at the start of the trial, the beer the randomisaon process has worked – note that it may not be possible for all aspects to be similar between groups, such as some treatment-specific dosing levels or regimens. □ □ Are the results presented and interpreted in line with the pre-defined stascal plan? Endpoints are oen analysed in a hierarchical manner, and the interpretaon of results should be consistent with this analysis plan – for example, secondary endpoints can only be stascally analysed or interpreted if the primary endpoint is met. □ □ Are esmates of effect size and precision given? For primary and secondary endpoints, an effect size esmate (such as Relave Risk [RR] or Number Needed to Treat [NNT]) should be given. Esmates of precision (such as confidence intervals [CIs]) should also be provided where appropriate – narrow CIs suggest a high level of confidence that the effect size esmate provided precisely reflects the true populaon effects size. CIs can also help to interpret stascal significance, as if a CI includes a value corresponding to no effect, then the effect size esmate is not stascally significant at the 0.05 level. □ □ Are adverse events reported? Detailed informaon about harms or unintended effects should be presented, as this may determine whether an intervenon is considered acceptable. □ □ DISCUSSION Are the results interpreted and discussed appropriately? Yes No The discussion secon should provide interpretaon that is consistent with the results presented, balancing benefits and harms and clearly addressing trial limitaons. Whether the results are generalisable to people outside of the trial should also be discussed. □ □ Fig. 1 Interpreting publications reporting results from randomised controlled trials (RCTs): a guidance checklist 2638 Diabetes Ther (2021) 12:2635–2644 used can introduce bias, including participants, has begun [15]. In this instance, it is important healthcare providers, data collectors, outcome that the authors are transparent regarding the adjudicators, and data analysts [15]. Unblinded changes and that the trial has been properly participants may act differently if they are aware blinded, because changing endpoints based on of their treatment group, while those involved unblinded data may introduce bias, compro- in trial organisation or data analysis may mising the reliability of the results. Such chan- unconsciously introduce bias through differen- ges may be less of a concern if the trial is an tial outcomes assessments and data analysis adaptive trial in which accumulating data are strategy choices. Ideally, neither the participant used to direct modifications to the endpoints in nor the treating physician would know which a manner that conforms to prespecified rules treatment was being received/given; this is [23]. known as a ‘double-blind’ study design [20]. A process known as allocation concealment Are All Primary and Secondary Endpoints should be employed where possible to prevent Clearly Defined? those managing participant treatment assign- Primary and secondary endpoints, as well as the ment from knowing the allocation in advance, statistical methodology used to analyse them, and is an important partner to robust ran- should be predefined as per the trial protocol domisation in preventing unintentional selec- [15]. Such definition and robust analysis plans tion bias [15]. Given the importance of being prevent the introduction of multiplicity and able to evaluate potential biases in an RCT, a subsequent false positives, whereby findings clear description of who was blinded and in stated as ‘significant’ occur by chance solely what manner should be provided. For example, because a large number of tests are performed. in a study of 40 consecutive RCT publications For example, if 100 tests are performed, five from five leading journals, eight different would be significant at a level of p \ 0.05 purely blinding combinations were employed in RCTs by chance. Well-defined endpoints also help to that reported being double blind [15, 21]. It is provide a suitable sample size estimation for a important to note that blinding is not always sufficiently powered study, which is usually possible in an RCT, and when outcomes are based on the primary endpoint [24]. Relevant objective (e.g. death), blinding is less important information should be clearly presented in the than if outcomes are subjective (e.g. symptoms) methods, and any analyses that were not pre- [20, 22]. defined should be clearly stated as ‘exploratory’ or ‘post hoc’. While studies can have multiple Was the Study Conducted in Line primary outcomes, this may complicate the with the Study Protocol? interpretation of a study drug’s efficacy unless Most trials will have a detailed protocol that multiplicity is clearly accounted for in the sta- specifies how the trial will be conducted. Devi- tistical analysis plan [15]. ating from the predefined protocol does not It is also worth noting that a clear descrip- necessarily mean that the study was conducted tion of primary and secondary endpoints, in incorrectly, but such deviations can affect the addition to other methodological details, can interpretation or relevance of a study and help to determine whether results should be should be clearly stated in the methods section compared with those from other RCTs. For along with the rationale [15, 17]. Protocol example, results from two head-to-head trials deviations can include failing to obtain the investigating the efficacy and safety of insulin desired number of study participants, changes glargine 300 U/mL versus insulin degludec to the inclusion and exclusion criteria, varia- 100 U/mL (IDeg) (the BRIGHT trial [10]) and tions in the provided treatments or interven- 200 U/mL (the CONCLUDE trial [11]) have tions, changes to the technologies used, and recently been published. Comparing the results changes to the duration of the follow-up [17]. of these two trials may seem logical, but the Even study endpoints, or the way in which primary endpoint of BRIGHT was related to they are assessed, may be changed after the trial HbA1c reduction whereas the rate of Diabetes Ther (2021) 12:2635–2644 2639 hypoglycaemia was the primary endpoint in aware that any losses to follow-up may have a CONCLUDE, so readers should carefully con- large impact on outcomes analyses in terms of sider this type of information before making introducing bias, especially if the number of direct comparisons. participants with the outcome under investiga- tion is small [20]. It should be noted that if Is the Sample Size Appropriate for the Analyses participants are excluded from the trial on the Being Done? grounds of protocol deviations, this should be Sample size is usually calculated based on the communicated in the text with an explanation ability to show a pre-defined and clinically rel- of the deviation [15]. evant between-group difference for the primary endpoint. Typically, the smaller the difference Were the Groups Similar at the Start is estimated to be, the larger the required par- of the Trial? ticipant population [15]. Statistically under- As previously stated, the aim of randomisation powered trials may still have some benefit in is to ensure equal division of any known or terms of downstream meta-analyses and if the unknown confounding factors present in the trial is unbiased, properly reported, and pub- participant population, meaning that any lished irrespective of results [25, 26]. However, effects on outcomes can be reliably attributed to there is the opposing view that underpowered the treatment [17]. Generally, the more similar studies have limited use unless for treatment of the comparator groups at the start of the trial, rare diseases in which plans for a prospective the better the randomisation process has meta-analysis are predefined or for trials in worked [20], although it should be noted that it earlier phases where randomised treatment may not be possible for all aspects to be similar comparisons is not the only defined goal [27]. between groups, due to factors such as treat- As with other study considerations, authors ment-specific dosing levels or regimens. While should clearly indicate how sample size was effective randomised assignment of participants determined in the methods, including the effect to treatment groups should theoretically size that the study is powered to detect. remove any selection bias, it is still possible for some differences between treatment groups to Results Section occur by chance, particularly when sample sizes are small [28]. For this reason, it is important to Are All Participants Accounted evaluate the baseline characteristics of the For in the Trial? comparator groups to determine whether there A trial should clearly show what happened to are any clinically important differences that each participant. This includes who received may need to be considered when evaluating the treatment, who was included in the analyses impact of a treatment on the study group. This (participants should ideally be analysed by the information is usually presented in a groups to which they were randomised, i.e. an table where values are supported by ranges or ‘intent-to-treat’ analysis), and who discontin- standard deviations of the mean where neces- ued the treatment [15]. This information should sary. Note that significance testing of these be captured in a flow diagram for clarity, and baseline characteristics between treatment discontinuations and exclusions should be groups is generally not advised and may be minimal [20]. It is important that the reason for misleading [15, 29]. Statistical testing between participant attrition is clearly indicated, as par- groups is a way of assessing whether any dif- ticipant loss to follow-up may have a very dif- ferences occur by chance, but if randomisation ferent impact on trial results to excluding has been performed correctly then any differ- specific participants because of treatment with- ences in baseline characteristics between the drawal or poor adherence, as the latter reasons treatment groups in an RCT are, by definition, are more likely to be related to the study treat- due to chance [29]. Instead, the reader of an ment [15]. Additionally, the reader should be RCT publication should consider whether any differences observed between groups are 2640 Diabetes Ther (2021) 12:2635–2644 clinically relevant and of a magnitude that may precision (such as confidence intervals [CIs]) impact the outcomes being assessed in the trial. where appropriate [15]—narrow CIs suggest a high level of confidence that the effect size Are the Results Presented and Interpreted estimate provided precisely reflects the true in Line with the Predefined Statistical Plan? population effect size. CIs can also help to Statistical testing should be predefined ahead of interpret statistical significance, as, in some data analysis in order to eliminate bias caused cases, if a 95% CI includes a value correspond- by choosing tests that may misconstrue the ing to no effect, then the effect size estimate is significance of the data, and such tests should not statistically significant at the 0.05 level [20]. be clearly described in the methodology section However, CIs are also useful when outcomes do [17]. An important point for readers to note is not meet the predefined statistical significance, that endpoints are often analysed in a hierar- in which case they may indicate that a clinically chical manner, and the interpretation of results important difference cannot be ruled out [15]. should be consistent with this analysis plan— The usefulness of effect size estimates and CIs for example, to avoid problems associated with is apparent when individual studies with nomi- multiplicity and so-called type 1 error, sec- nally nonsignificant results are subsequently ondary endpoints can only be statistically included in meta-analyses pooling multiple analysed or interpreted if the primary endpoint studies. For example, in both the EDITION 3 is met [30, 31]. An example of this is demon- (insulin glargine 300 U/mL [Gla-300] versus strated in the recent SoliMix trial comparing the insulin glargine 100 U/mL) and BEGIN Once fixed-ratio combination of iGlarLixi (basal Long (insulin degludec versus insulin glargine insulin glargine 100 U/mL plus the glucagon- 100 U/mL) RCTs in people with type 2 diabetes, like peptide-1 receptor agonist lixisenatide) 95% CIs objectively suggested no difference with the premix insulin analogue BIAsp 30 between the treatment arms in confirmed (30% insulin aspart and 70% insulin aspart hypoglycaemia occurring at any time of day protamine) [32]. A primary endpoint of this (based on analyses of incidence in EDITION 3 study was the noninferiority of iGlarLixi to and rates in BEGIN Once Long) [1, 9]. However, BIAsp 30 in HbA1c reduction from baseline to in both cases, the upper CI only crossed 1 (which week 26. Superiority of iGlarLixi to BIAsp 30 in indicated no effect) by a small margin, as the RR HbA1c reduction from baseline to week 26 was was 0.88 (95% CI 0.77–1.01) in EDITION 3 and also an endpoint; however, it was a hierarchical the rate ratio was 0.82 (0.64–1.04) in BEGIN Once key secondary endpoint that would only be Long [1, 9]. Later, meta-analyses including these interpreted if the primary endpoint and the key studies would show that both Gla-300 and IDeg secondary endpoints before it in the hierarchy provided a benefit over insulin glargine 100 U/mL were successfully met. When a primary end- in terms of hypoglycaemia, with similar effect point does not achieve statistical significance, sizes to the individual studies but narrower CIs: secondary endpoints may still be of interest and 0.91 (0.87–0.96) and 0.83 (0.74–0.94), are useful for generating hypotheses [33], but respectively [34, 35]. readers should reflect that results for these sec- ondary endpoints should be considered Are Adverse Events Reported? exploratory and interpreted with some caution. Understanding the potentially harmful effects of a treatment is as important to a reader as Are Estimates of Effect Size and Precision understanding the positive benefits the treat- Given? ment may offer [15]. The analysis of certain For primary and secondary endpoints, an esti- events, such as hypoglycaemia in RCTs of dia- mate of the between-group difference (or effect betes therapies, will be captured within the size) such as relative risk (RR) or number needed primary and secondary endpoints and should to treat should be given [20]. These values be reported as described above. However, all should be supplemented with estimates of adverse events should be closely monitored and clearly reported in the results section of the RCT Diabetes Ther (2021) 12:2635–2644 2641 publication, typically using a detailed table to other information should be provided that may break down the adverse events experienced and reassure the reader that the authors (and, if the proportion of the participants who experi- applicable, the sponsor) are committed to high- enced them in each treatment group [36]. The quality, transparent scientific communication. number needed to harm may also be reported Clinical trial registration is important to [37]. The data recorded during an RCT con- ensure that there is sufficient transparency cerning harms or unintended side effects is key around how a clinical trial is conducted and to determining whether an intervention is that data reporting is not biased or selective deemed acceptable for use [15]. The publication [15, 38, 39]. Moreover, the International Com- should report the criteria for event confirma- mittee of Medical Journal Editors (ICMJE) rec- tion, method of data collection, number of ommends that journals require clinical trials to participants withdrawn due to adverse events be registered in a public trials registry at or (whether treatment-related or otherwise), and before the time of first patient enrolment in the absolute risk of each event, where possible order to be considered for publication [39]. [36]. Registration in these databases requires the provision of key information prescribed by a Discussion Section 24-point checklist, including but not limited to primary and secondary outcomes, completion Are the Results Interpreted and Discussed dates, ethics, sponsor information, interven- Appropriately? tions, study type, and sample size [39]. As such, The discussion section of an RCT publication clinical trial registry pages can be a valuable should provide interpretation that is consistent source of key information that may not all be with the results presented, balancing benefits available within associated publications. Publi- and harms and considering the importance of cations reporting results from a clinical trial the findings in view of the current published without a registration number in a rep- evidence (some of which will have been pre- utable registry system should be considered sented in the introduction) [15]. The main focus carefully. should be on the primary endpoint, as this is Publications will typically include a list of the outcome considered to be the most impor- potential conflicts of interest (COIs) for the tant in the context of the study [24]. Secondary authors of the article. This is essential for com- endpoints are discussed also, albeit with less plete transparency regarding any potential bia- importance, and exploratory endpoints, while ses or influences (e.g. financial gain) that may not typically a focus, can be included for influence the work [40]. It is important to note introducing new hypotheses [30]. Whether the that the presence of a potential COI does not results are generalisable to people outside of the necessarily indicate that the results being pre- trial should also be discussed, as RCT inclusion sented are compromised, and it is up to the criteria are often restrictive [15]. Finally, while reader to evaluate these conflicts and determine often excluded, it is important for RCTs to whether they believe there to be a COI that clearly address any trial limitations that may impacts the credibility of the results. affect the results [15]. These include, but are not limited to, the available participant population, CONCLUSION the type of study being performed, and whether there were any study design considerations. RCTs are extremely robust when conducted and reported appropriately, but determining whe- Other Information Provided ther this has been done and assessing the validity of RCT results can be confusing for While the sections above comprise what we individuals who may be unfamiliar with their believe to be the key aspects to consider when interpretation. Guidelines regarding RCTs are interpreting the content of an RCT publication, specific in their intended use. For example, 2642 Diabetes Ther (2021) 12:2635–2644 CONSORT is a guide to reporting the results of reviewing drafts of the manuscript and RCTs, while developing the trial protocol itself approved the final version for submission. is covered by the SPIRIT (Standard Protocol Items: Recommendations for Interventional Disclosures. Tadej Battelino has received Trials) guidelines [41]. These two guidelines honoraria for participation on advisory boards may have some overlapping content but they for Novo Nordisk, Sanofi, Eli Lilly, Boehringer are designed with different target audiences in Ingelheim, and Medtronic, and as a speaker for mind. Similarly, the checklist presented in this AstraZeneca, Eli Lilly, Novo Nordisk, Medtronic, commentary has been developed primarily for Pfizer, Sanofi, Dexcom, Abbott and Roche; owns clinicians who will be reading RCT publications. stocks of DreaMed Diabetes; T.B.’s institution It distills the extensive information provided by has received research grant support and travel CONSORT, in addition to other published lit- expenses from Abbott Diabetes Care, Medtronic, erature and our own experience, into a refined Novo Nordisk, GluSense, Sanofi, Sandoz, checklist of key points specifically focusing on Novartis and Zealand. Didac Mauricio has the interpretation of results from RCTs (Fig. 1). served on advisory panels for Ferrer, Sanofi, This checklist can be used alone or in conjunc- Novo Nordisk, Janssen and MSD, and received tion with other aids such as CONSORT to help speakers’ bureau fees from Almirall, Boehringer HCPs evaluate the quality of evidence provided Ingelheim, Esteve, Eli Lilly, Janssen, Sanofi, by RCT publications. Novo Nordisk, MSD and Urgo. Institutional In summary, with the checklist provided, research support was obtained from AstraZeneca, and the explanation of why we consider the GlaxoSmithKline, Eli Lilly, Sanofi, Novo Nordisk points discussed above to be worthy of atten- and MSD. tion, we hope that this article will provide a useful guide for HCPs when reading and inter- Compliance with Ethics Guidelines. This preting publications that report results of RCTs. article is based on previously conducted studies and does not contain any new studies with human participants or animals performed by any of the authors. ACKNOWLEDGEMENTS Data Availability. Data sharing is not applicable to this article as no datasets were Funding. Funding for the development of generated or analysed during the current study. this manuscript, including editorial assistance and the journal’s Rapid Service Fee, was pro- Open Access. This article is licensed under a vided by Sanofi (Paris, France). Creative Commons Attribution-NonCommer- cial 4.0 International License, which permits Editorial Assistance. Editorial assistance any non-commercial use, sharing, adaptation, was provided by Simon Rees, PhD, and Cam distribution and reproduction in any medium Hubert, PhD, of Fishawack Communications or format, as long as you give appropriate credit Ltd, part of Fishawack Health and funded by to the original author(s) and the source, provide Sanofi. a link to the Creative Commons licence, and indicate if changes were made. The images or Authorship. All named authors meet the other third party material in this article are International Committee of Medical Journal included in the article’s Creative Commons Editors (ICMJE) criteria for authorship for this licence, unless indicated otherwise in a credit article, take responsibility for the integrity of line to the material. If material is not included the work as a whole, and have given their in the article’s Creative Commons licence and approval for this version to be published. your intended use is not permitted by statutory Authors’ Contributions. All authors were regulation or exceeds the permitted use, you involved in the data interpretation, writing and will need to obtain permission directly from the Diabetes Ther (2021) 12:2635–2644 2643 copyright holder. To view a copy of this licence, (BEGIN: Once Simple Use). Adv Ther. 2013;30(6): visit http://creativecommons.org/licenses/by- 607–22. nc/4.0/. 9. Zinman B, et al. Insulin degludec versus insulin glargine in insulin-naive patients with type 2 dia- betes: a 1-year, randomized, treat-to-target trial (BEGIN Once Long). Diabetes Care. 2012;35(12): REFERENCES 2464–71. 10. Rosenstock J, et al. More similarities than differ- 1. Bolli GB, et al. New insulin glargine 300 U/ml ences testing insulin glargine 300 units/mL versus compared with glargine 100 U/ml in insulin-naive insulin degludec 100 units/mL in insulin-naive type people with type 2 diabetes on oral glucose-lower- 2 diabetes: the randomized head-to-head BRIGHT ing drugs: a randomized controlled trial (EDITION trial. Diabetes Care. 2018;41(10):2147–54. 3). Diabetes Obes Metab. 2015;17(4):386–94. 11. Philis-Tsimikas A, et al. Risk of hypoglycaemia with 2. Riddle MC, et al. New insulin glargine 300 units/mL insulin degludec versus insulin glargine U300 in versus glargine 100 units/mL in people with type 2 insulin-treated patients with type 2 diabetes: the diabetes using basal and mealtime insulin: glucose randomised, head-to-head CONCLUDE trial. Dia- control and hypoglycemia in a 6-month random- betologia. 2020;63(4):698–710. ized controlled trial (EDITION 1). Diabetes Care. 2014;37(10):2755–62. 12. Davies M, et al. Initiation of insulin glargine ther- apy in type 2 diabetes subjects suboptimally con- 3. Terauchi Y, et al. New insulin glargine 300 U/ml trolled on oral antidiabetic agents: results from the versus glargine 100 U/ml in Japanese people with AT.LANTUS trial. Diabetes Obes Metab. 2008;10(5): type 2 diabetes using basal insulin and oral anti- 387–99. hyperglycaemic drugs: glucose control and hypo- glycaemia in a randomized controlled trial 13. Hermansen K, et al. A 26-week, randomized, paral- (EDITION JP 2). Diabetes Obes Metab. 2016;18(4): lel, treat-to-target trial comparing insulin detemir 366–74. with NPH insulin as add-on therapy to oral glucose- lowering drugs in insulin-naive people with type 2 4. Bode BW, et al. Efficacy and safety of insulin diabetes. Diabetes Care. 2006;29(6):1269–74. degludec 200 U/mL and insulin degludec 100 U/mL in patients with type 2 diabetes (BEGIN: Compare). 14. Riddle MC, Rosenstock J, Gerich J. The treat-to- Endocr Pract. 2014;20(8):785–91. target trial: randomized addition of glargine or human NPH insulin to oral therapy of type 2 dia- 5. Yki-Järvinen H, et al. New insulin glargine 300 betic patients. Diabetes Care. 2003;26(11):3080–6. units/mL versus glargine 100 units/mL in people with type 2 diabetes using oral agents and basal 15. Moher D, et al. CONSORT 2010 explanation and insulin: glucose control and hypoglycemia in a elaboration: updated guidelines for reporting par- 6-month randomized controlled trial (EDITION 2). allel group randomised trials. BMJ. 2010;340:c869. Diabetes Care. 2014;37(12):3235–43. 16. World Medical Association. World Medical Associ- 6. Garber AJ, et al. Insulin degludec, an ultra-lon- ation Declaration of Helsinki: ethical principles for gacting basal insulin, versus insulin glargine in medical research involving human subjects. JAMA. basal-bolus treatment with mealtime insulin aspart 2013;310(20):2191–4. in type 2 diabetes (BEGIN Basal-Bolus Type 2): a phase 3, randomised, open-label, treat-to-target 17. Young JM, Solomon MJ. How to critically appraise non-inferiority trial. Lancet. 2012;379(9825): an article. Nat Rev Gastroenterol Hepatol. 1498–507. 2009;6(2):82–91. 7. Gough SC, et al. Low-volume insulin degludec 200 18. National Health and Medical Research Council. units/ml once daily improves glycemic control How to review the evidence: systematic identifica- similarly to insulin glargine with a low risk of tion and review of the scientific literature. Can- hypoglycemia in insulin-naive patients with type 2 berra: National Health and Medical Research diabetes: a 26-week, randomized, controlled, Council; 2000. multinational, treat-to-target trial: the BEGIN LOW VOLUME trial. Diabetes Care. 2013;36(9):2536–42. 19. Dumville JC, et al. The use of unequal randomisa- tion ratios in clinical trials: a review. Contemp Clin 8. Philis-Tsimikas A, et al. Insulin degludec once-daily Trials. 2006;27(1):1–12. in type 2 diabetes: simple or step-wise titration 2644 Diabetes Ther (2021) 12:2635–2644 20. The Centre for Evidence-Based Medicine. RCT crit- europa.eu/en/documents/scientific-guideline/draft- ical appraisal sheet. https://www.cebm.net/2014/ guideline-multiplicity-issues-clinical-trials_en.pdf. 06/critical-appraisal/. Accessed 4 Dec 2020. Accessed Dec 2019. 21. Montori VM, et al. In the dark: the reporting of 32. McCrimmon RJ, et al. Advancing therapy with blinding status in randomized controlled trials. iGlarLixi versus premix BIAsp 30 in basal insulin- J Clin Epidemiol. 2002;55(8):787–90. treated type 2 diabetes: design and baseline char- acteristics of the SoliMix randomized controlled 22. Bolli GB, et al. Glycaemic control and hypogly- trial. Diabetes Obes Metab. 2021;23(6):1221–31. caemia with insulin glargine 300U/mL versus insulin glargine 100U/mL in insulin-naive people 33. Pocock SJ, Stone GW. The primary outcome fails— with type 2 diabetes: 12-month results from the what next? N Engl J Med. 2016;375(9):861–70. EDITION 3 trial. Diabetes Metab. 2017;43(4):351–8. 34. Ritzel R, et al. Patient-level meta-analysis of the 23. Pallmann P, et al. Adaptive designs in clinical trials: EDITION 1, 2 and 3 studies: glycaemic control and why use them, and how to run and report them. hypoglycaemia with new insulin glargine 300 U/ml BMC Med. 2018;16(1):29. versus glargine 100 U/ml in people with type 2 diabetes. Diabetes Obes Metab. 2015;17(9):859–67. 24. Andrade C. The primary outcome measure and its importance in clinical trials. J Clin Psychiatry. 35. Ratner RE, et al. Hypoglycaemia risk with insulin 2015;76(10):e1320–3. degludec compared with insulin glargine in type 2 and type 1 diabetes: a pre-planned meta-analysis of 25. Guyatt GH, Mills EJ, Elbourne D. In the era of sys- phase 3 trials. Diabetes Obes Metab. 2013;15(2): tematic reviews, does the size of an individual trial 175–84. still matter? PLoS Med. 2008;5(1):e4. 36. Ioannidis JP, et al. Better reporting of harms in 26. Schulz KF, Grimes DA. Sample size calculations in randomized trials: an extension of the CONSORT randomised trials: mandatory and mystical. Lancet. statement. Ann Intern Med. 2004;141(10):781–8. 2005;365(9467):1348–53. 37. Sedgwick P. What is number needed to harm 27. Halpern SD, Karlawish JH, Berlin JA. The continu- (NNH)? BMJ Br Med J. 2013;347:f4869. ing unethical conduct of underpowered clinical trials. JAMA. 2002;288(3):358–62. 38. Aslam A, et al. Registration of clinical trials: is it really needed? N Am J Med Sci. 2013;5(12):713–5. 28. Lewis SC, Warlow CP. How to spot bias and other potential problems in randomised controlled trials. 39. ICMJE. Clinical trial recommendations—registra- J Neurol Neurosurg Psychiatry. 2004;75(2):181–7. tion. 2021. https://prsinfo.clinicaltrials.gov/ trainTrainer/WHO-ICMJE-ClinTrialsgov-Cross-Ref. 29. de Boer MR, et al. Testing for baseline differences in pdf. Accessed 9 April 2021. randomized controlled trials: an unhealthy research behavior that is hard to eradicate. Int J Behav Nutr 40. ICMJE. Recommendations—disclosure of financial Phys Act. 2015;12:4. and non-financial relationships and activities, and conflicts of interest. 2021. http://www.icmje.org/ 30. US Food and Drug Administration (FDA). Guidance recommendations/browse/roles-and- for industry, multiple endpoints in clinical trials responsibilities/author-responsibilities–conflicts-of- (January 2017 draft). https://www.fda.gov/media/ interest.html. Accessed 9 April 2021. 102657/download. Accessed Dec 2019. 41. Chan AW, et al. SPIRIT 2013 statement: defining 31. European Medicines Agency. Guideline on multi- standard protocol items for clinical trials. Ann plicity issues in clinical trials (EMA/CHMP/44762/ Intern Med. 2013;158(3):200–7. 201, December 2016 draft). https://www.ema.