- Email: [email protected]
Preregistration of study design and non-inferiority margin
Correspondence
Institute of Population Health Sciences, National Health Research Institutes, Zhunan, Taiwan (CPW, MKT); China Medical University Hospital, Taichung, Taiwan (CPW, MKT); Laboratory for Exercise Physiology Research, Institute of Sport Science, National Taiwan Sport University, Taoyuan, Taiwan (JPMW); and Department of Epidemiology, MD Anderson Cancer Center, Houston, TX, USA (XW) 1
2
3 4
5
Wen CP, Wu X. Stressing harms of physical inactivity to promote exercise. Lancet 2012; 380: 192–93. WHO. WHO report on the global tobacco epidemic, 2008: the MPOWER package. Geneva: World Health Organization; 2008. http://www.who.int/tobacco/mpower/2008/ en/index.html (accessed Dec 18, 2012). Norcross JC, Krebs PM, Prochaska JO. Stages of change. J Clin Psychol 2011; 67: 143–54. Tuah NA, Amiel C, Qureshi S, Car J, Kaur B, Majeed A. Transtheoretical model for dietary and physical exercise modification in weight loss management for overweight and obese adults. Cochrane Database Syst Rev 2011; 10: CD008066. Chapman S, Freeman B. Markers of the denormalisation of smoking and the tobacco industry. Tob Control 2008; 17: 25–31.
safety, and tolerability of linagliptin compared with glimepiride. Neither the non-inferiority design, nor the margin of 0·35%, was mentioned. Moreover, of the 27 secondary outcome measures, 26 were added to ClinicalTrials.gov on Feb 17, 2012, whereas the trial took place between Feb 12, 2008, and Dec 12, 2010. In our view, it is essential to prespecify the exact study design and the non-inferiority margin, and to publish that information, preferably before the study starts. When this information is not in the trial registry, the study protocol should be published and referred to in the paper. We declare that we have no conflicts of interest.
*Kornelis J J van Hateren, Nanne Kleefstra, Henk J Bilo [email protected]
Preregistration of study design and noninferiority margin Baptist Gallwitz and colleagues (Aug 4, p 475)1 conclude that linagliptin is noninferior to glimepiride in lowering glycated haemoglobin (HbA1c), and is associated with fewer cardiovascular events. Non-inferiority trials are not designed to show any additional value of new drugs; they are just aimed at showing whether a new treatment has as much efficacy as another treatment or performs worse. There has been much debate about this type of study and by some it is regarded as unethical because it seems to disregard patients’ interests.2 A failure to prespecify primary (and secondary) endpoints can introduce bias and creates opportunities for manipulation.3 In non-inferiority studies, it is important to prespecify the non-inferiority margin. Also, such information should be preregistered. According to the registry at ClinicalTrials.gov (NCT00622284), the objective of Gallwitz and colleagues’ trial was to investigate the efficacy, www.thelancet.com Vol 381 January 12, 2013
Diabetes Centre, Isala Clinics, PO Box 10400, Zwolle, 8000 GK Overijssel, Netherlands 1
2
3
Gallwitz B, Rosenstock J, Rauch T, et al. 2-year efficacy and safety of linagliptin compared with glimepiride in patients with type 2 diabetes inadequately controlled on metformin: a randomised, double-blind, noninferiority trial. Lancet 2012; 380: 475–83. Garattini S, Bertele’ V. Non-inferiority trials are unethical because they disregard patients’ interests. Lancet 2007; 370: 1875–77. Evans S. When and how can endpoints be changed after initiation of a randomized clinical trial? PLoS Clin Trials 2007; 2: e18.
Authors’ reply We disagree with Kornelis van Hateren and colleagues that trials with a non-inferiority design can be regarded as unethical. The design of our trial1 and the primary endpoint of change in glycated haemoglobin (HbA1c) was chosen on the basis of guidance provided by regulatory authorities for development of diabetes drugs2,3 and statistical principles of the International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use.4 As described by the US Food and Drug Administration (FDA)2 and the European Medicines Agency (EMA),3 efficacy trials should aim to show non-inferiority of the new agent to
an active comparator with respect to change in HbA1c (in addition to showing superiority over placebo in separate studies). Furthermore, we also disagree that a non-inferiority trial cannot show any additional value of a new drug, since this design does not preclude exploration of other potential benefits via secondary endpoints. In fact, the EMA guidance3 describes the importance of balancing the margin of potential inferiority against possible clinical advantages, including tolerability and cardiovascular risk profile—outcomes which we assessed in our study by doing prespecified inferential statistical analyses of hypoglycaemia, bodyweight, and cardiovascular events. Indeed, the cardiovascular hazard ratio for linagliptin versus glimepiride in our study provided a hypothesis-generating signal, which is currently being tested in a large cardiovascular outcomes study (CAROLINA; NCT01243424). Van Hateren and colleagues incorrectly imply that the non-inferiority design and study endpoints were not prespecified. The full study protocol— including a complete description of the non-inferiority design and margin, and all the study endpoints (primary, secondary, and exploratory)—was finalised in October, 2007, after review and approval by regulatory authorities, and the first patient was subsequently enrolled in February, 2008. It is clear from the FDA’s publicly available medical review of linagliptin that our study’s protocol was prespecified and approved by regulatory authorities.5 Moreover, the study protocol was submitted to The Lancet together with our paper and was available for peer review. ClinicalTrials.gov provides summary detail rather than a complete repository of all information pertaining to clinical studies, and there is no place to capture the element of non-inferiority design. Indeed, a preliminary search of 115
Institute of Population Health Sciences, National Health Research Institutes, Zhunan, Taiwan (CPW, MKT); China Medical University Hospital, Taichung, Taiwan (CPW, MKT); Laboratory for Exercise Physiology Research, Institute of Sport Science, National Taiwan Sport University, Taoyuan, Taiwan (JPMW); and Department of Epidemiology, MD Anderson Cancer Center, Houston, TX, USA (XW) 1
2
3 4
5
Wen CP, Wu X. Stressing harms of physical inactivity to promote exercise. Lancet 2012; 380: 192–93. WHO. WHO report on the global tobacco epidemic, 2008: the MPOWER package. Geneva: World Health Organization; 2008. http://www.who.int/tobacco/mpower/2008/ en/index.html (accessed Dec 18, 2012). Norcross JC, Krebs PM, Prochaska JO. Stages of change. J Clin Psychol 2011; 67: 143–54. Tuah NA, Amiel C, Qureshi S, Car J, Kaur B, Majeed A. Transtheoretical model for dietary and physical exercise modification in weight loss management for overweight and obese adults. Cochrane Database Syst Rev 2011; 10: CD008066. Chapman S, Freeman B. Markers of the denormalisation of smoking and the tobacco industry. Tob Control 2008; 17: 25–31.
safety, and tolerability of linagliptin compared with glimepiride. Neither the non-inferiority design, nor the margin of 0·35%, was mentioned. Moreover, of the 27 secondary outcome measures, 26 were added to ClinicalTrials.gov on Feb 17, 2012, whereas the trial took place between Feb 12, 2008, and Dec 12, 2010. In our view, it is essential to prespecify the exact study design and the non-inferiority margin, and to publish that information, preferably before the study starts. When this information is not in the trial registry, the study protocol should be published and referred to in the paper. We declare that we have no conflicts of interest.
*Kornelis J J van Hateren, Nanne Kleefstra, Henk J Bilo [email protected]
Preregistration of study design and noninferiority margin Baptist Gallwitz and colleagues (Aug 4, p 475)1 conclude that linagliptin is noninferior to glimepiride in lowering glycated haemoglobin (HbA1c), and is associated with fewer cardiovascular events. Non-inferiority trials are not designed to show any additional value of new drugs; they are just aimed at showing whether a new treatment has as much efficacy as another treatment or performs worse. There has been much debate about this type of study and by some it is regarded as unethical because it seems to disregard patients’ interests.2 A failure to prespecify primary (and secondary) endpoints can introduce bias and creates opportunities for manipulation.3 In non-inferiority studies, it is important to prespecify the non-inferiority margin. Also, such information should be preregistered. According to the registry at ClinicalTrials.gov (NCT00622284), the objective of Gallwitz and colleagues’ trial was to investigate the efficacy, www.thelancet.com Vol 381 January 12, 2013
Diabetes Centre, Isala Clinics, PO Box 10400, Zwolle, 8000 GK Overijssel, Netherlands 1
2
3
Gallwitz B, Rosenstock J, Rauch T, et al. 2-year efficacy and safety of linagliptin compared with glimepiride in patients with type 2 diabetes inadequately controlled on metformin: a randomised, double-blind, noninferiority trial. Lancet 2012; 380: 475–83. Garattini S, Bertele’ V. Non-inferiority trials are unethical because they disregard patients’ interests. Lancet 2007; 370: 1875–77. Evans S. When and how can endpoints be changed after initiation of a randomized clinical trial? PLoS Clin Trials 2007; 2: e18.
Authors’ reply We disagree with Kornelis van Hateren and colleagues that trials with a non-inferiority design can be regarded as unethical. The design of our trial1 and the primary endpoint of change in glycated haemoglobin (HbA1c) was chosen on the basis of guidance provided by regulatory authorities for development of diabetes drugs2,3 and statistical principles of the International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use.4 As described by the US Food and Drug Administration (FDA)2 and the European Medicines Agency (EMA),3 efficacy trials should aim to show non-inferiority of the new agent to
an active comparator with respect to change in HbA1c (in addition to showing superiority over placebo in separate studies). Furthermore, we also disagree that a non-inferiority trial cannot show any additional value of a new drug, since this design does not preclude exploration of other potential benefits via secondary endpoints. In fact, the EMA guidance3 describes the importance of balancing the margin of potential inferiority against possible clinical advantages, including tolerability and cardiovascular risk profile—outcomes which we assessed in our study by doing prespecified inferential statistical analyses of hypoglycaemia, bodyweight, and cardiovascular events. Indeed, the cardiovascular hazard ratio for linagliptin versus glimepiride in our study provided a hypothesis-generating signal, which is currently being tested in a large cardiovascular outcomes study (CAROLINA; NCT01243424). Van Hateren and colleagues incorrectly imply that the non-inferiority design and study endpoints were not prespecified. The full study protocol— including a complete description of the non-inferiority design and margin, and all the study endpoints (primary, secondary, and exploratory)—was finalised in October, 2007, after review and approval by regulatory authorities, and the first patient was subsequently enrolled in February, 2008. It is clear from the FDA’s publicly available medical review of linagliptin that our study’s protocol was prespecified and approved by regulatory authorities.5 Moreover, the study protocol was submitted to The Lancet together with our paper and was available for peer review. ClinicalTrials.gov provides summary detail rather than a complete repository of all information pertaining to clinical studies, and there is no place to capture the element of non-inferiority design. Indeed, a preliminary search of 115