ICTW Cordoba, Argentina04/11/2015 Publishing (and presenting) the results of your trial Ian F Tannock MD, PhD, DSc Professor of Medical Oncology and Medical Biophysics Princess Margaret Cancer Centre and University of Toronto

Aims of the presentation To understand: The ethical obligation to publish - and the problem of Publication Bias Principles of authorship What to include in an abstract - both for a publication and presentation at a meeting What to include in a paper What to avoid in a paper 04/11/2015ICTW Cordoba, Argentina

04/11/2015ICTW Cordoba, Argentina Publication is an ethical requirement Research has no value unless its results are available to others. Failure to publish breaks an implicit contract with the patients who participated, and with those who funded the research. You have an ethical obligation to publish the results of your research

04/11/2015ICTW Cordoba, Argentina Improvements in Outcome... Can only be built on evidence, and not on opinion “The opinion of experts has been a traditional source of all the errors throughout medical history” Alvin Feinstein You are building that evidence by performing and publishing your clinical trial

04/11/2015ICTW Cordoba, Argentina Publication (and Presentation) Bias Publication should depend on the quality of the study Authors are more likely to publish or present a study if it appears to have “positive” results Journals and Meetings are more likely to accept a paper if it appears to have “positive” results This leads to bias in the literature, and can lead to inappropriate management of patients

04/11/2015ICTW Cordoba, Argentina Publication Bias applies even to large randomized trials Krzyzanowska et al JAMA 2003;290:495-50) For 510 RCTs reported in ASCO abstracts, there was delayed publication of negative trials 68% -ve studies 81% +ve studies

04/11/2015ICTW Cordoba, Argentina

04/11/2015ICTW Cordoba, Argentina Failure to publish important negative studies…. … was more often due to investigators losing interest than journals refusing to publish a “negative” paper Make sure you publish your results, regardless of the results of your study!

Developing principles for authorship Principles of authorship should be decided before a trial is opened. All authors must have made substantive contributions to designing, undertaking, analyzing and/or reporting the study The chair of the study is responsible for drafting the main MS describing its results Other authors would usually include –Study statistician –Investigators active on the steering committee –Local investigators who accrue more than a predefined number of patients 04/11/2015ICTW Cordoba, Argentina

Avoid “ghost” and “honorary” authors A ghostwriter writes articles that are officially credited to another person Medical writers employed by companies who often are encouraged to write papers that are biased to the commercial goals An honorary author contributes minimally to the study International opinion leaders who may be offered lead authorship by sponsors to encourage adoption of the new treatment 04/11/2015ICTW Cordoba, Argentina

Sharing the data Data should be held centrally until the trial is completed For phase 3 RCTs no investigator should have access to the data prior to completion of the trial Safety and futility should be considered by an independent Data Monitoring Committee (DMC) All data must be freely available to the chair and steering committee at trial completion Data may be made available to others for secondary analysis only after the primary publication 04/11/2015ICTW Cordoba, Argentina

How to write a clear and concise scientific abstract and paper 04/11/2015ICTW Cordoba, Argentina

What should be included in the abstract of your paper or presentation? 1. The primary endpoint should be defined explicitly in the abstract of the paper or presentation 2. The concluding sentence of the abstract should relate only to the primary endpoint 3. The abstract should include a statement of major toxicity Unfortunately many reports, both in high level journals and at the ASCO meeting, do NOT meet these simple requirements 04/11/2015 ICTW Cordoba, Argentina

04/11/2015ICTW Cordoba, Argentina Recommendations for what to include when reporting a trial in an abstract Rationale for the study (52%) Brief description of the intervention (88%) Explicit definition of primary endpoint (22%) Magnitude of difference for primary endpoint with confidence limits or 2-sided p-value (97%) Sample size and brief description of participants (40%) Planned sample size or power of study (2%) Duration of follow-up (23%) Description of major toxicity (55%) Source of funding (17%)

04/11/2015ICTW Cordoba, Argentina Recommendations for what not to include in an abstract Analyses of endpoints that are not pre- specified in the protocol Results of subgroup analysis that are not pre-specified in the protocol This is data dredging and can give results that are statistically invalid

What should be included in your paper? 04/11/2015ICTW Cordoba, Argentina

04/11/2015ICTW Cordoba, Argentina CONSORT statement of what to include in a report of a clinical trial MethodsParticipants Interventions Objectives Outcomes Sample size Randomisation Blinding Statistical methods ResultsParticipant flow Recruitment Baseline data Numbers analysed Outcomes and estimation Ancillary analyses Adverse events DiscussionInterpretation Generalisability Overall evidence

04/11/2015ICTW Cordoba, Argentina If you use this type of checklist for your abstracts or papers You will avoid missing important items in describing the design and results of your clinical trial

04/11/2015ICTW Cordoba, Argentina Make sure that your main analysis applies to the pre-defined primary endpoint of your study Recognize also the limitations of the endpoint(s) that you use

We identified 164 reports of RCTs for breast cancer, published between 1995 and of 164 trials showed no significant difference in their primary endpoint In 59% of reports of these 92 trials the concluding statement of the abstract used secondary endpoints to suggest benefit Only one third of reports indicated the frequency of grade 3-4 toxicity in the abstract 04/11/2015ICTW Cordoba, Argentina

04/11/2015ICTW Cordoba, Argentina Factors leading to false-positive trials 1. Publication bias 2. Low probability that a new treatment will be superior 3. Multiple significance tests “Could the results of the trial be falsely positive?”

04/11/2015ICTW Cordoba, Argentina Multiple Significance Tests (Data Torturing) They are used in comparing treatment effects between randomized groups due to: 1. Use of multiple endpoints 2. Subgroup analysis 3. Repeated analysis of data after different time intervals

04/11/2015ICTW Cordoba, Argentina Beware of Subset Analysis 5-FU and levamisole as adjuvant treatment for Dukes C colon cancer 1. Mayo Clinic Trial (Laurie et al, JCO 1989;7:1447) More effective for men, older patients 2. Southwest Oncology Group Trial (Moertel et al, NEJM 1990;322:352) More effective for women, younger patients

04/11/2015ICTW Cordoba, Argentina What to avoid in your paper Length Conclusions not based on the primary endpoint Multiple significance tests & analysis of subgroups Overstatement Claiming to be first Criticism of others Comparing with other single arm studies

Tannock’s rules for posters You must stand 3 meters (10 feet) away from it and be able to read every word. Hence: –Minimum font size is 24 or 20 bold –Do not include an abstract that nobody can read – you can provide an abstract as a hand-out if you wish –Use point-form and not long sentences You can make at most 3 main points –The more you put in the less people will remember 04/11/2015ICTW Cordoba, Argentina

04/11/2015ICTW Cordoba, Argentina Make sure that the report reflects the results of the trial Beware of concluding statements that do not apply to the primary endpoint … “Temazolamide demonstrates efficacy equal to that of DTIC and is an alternative for patients with advanced malignant melanoma” Publication in JCO 2000 … or do not reflect the results obtained … or inflate effects favoured by the sponsor “ Temazolamide demonstrates inefficacy equal to that of DTIC and neither drug can be recommended for patients with melanoma” My interpretation

Acknowledgements My thanks to many present and former fellows, who undertook much of the work presented, and stimulated many of the ideas that led to it. Especially: Eitan Amir Chris Booth Alberto Ocana Bostjan Seruga Arnoud Templeton ‘Paco’ Vera-Badillo November 4, 2015 NCIC Cosbie Lecture

And now, 2 multiple choice questions 04/11/2015ICTW Cordoba, Argentina

Publication Bias: A.Is the bias toward concluding that results are better than found by objective measures B.Is the bias to publishing positive results more often than negative results C.Is rare for reports of phase III randomised controlled trials D.Leads to the impression that a new treatment is better than it really is E.Options A & C are correct F.Options B & D are correct 04/11/2015ICTW Cordoba, Argentina

You complete a single arm phase II study evaluating a new treatment in 30 patients. Your paper describing the trial should include: A.Analysis of outcome by age, gender and performance status B.A concluding statement based on the primary endpoint C.A concluding statement about effects on tumour response, time to progression and overall survival D.A statistical comparison of results with other phase II trials for the same disease E.A statement about the role of the new treatment in routine patient management 04/11/2015ICTW Cordoba, Argentina