Posted on behalf of Marcia Levenstein, Pfizer.Between 2013 and 2016 the committee on Professional Ethics updated ASA’s
Ethical Guidelines for Statistical Practice. The previous version had been approved in 1999 and advances in statistical practice, proliferation of data sources and analysts, and increased visibility of statistical results, (e.g. election predictions, use of big data), led to the recognition that the Guidelines should be revised to reflect the current landscape.
The ethical guidelines that an organization sets can determine how that profession is viewed by society as well as by its practitioners. Therefore, a good set of guidelines establishes a foundation to help our profession – but only if practitioners are familiar with them. The objective of the Guidelines is to assist statisticians, and practitioners of statistics from other fields, to make decisions ethically in the course of their work. The principles form a framework to guide our behavior, not a rigid set of rules to follow.
We are surrounded by codes of conduct in our workplaces, so these Ethical Guidelines shouldn’t be applied in isolation, but need to be interpreted and applied in the context of a particular situation-which might include consideration of the ethical guidelines from other disciplines or professions. The ASA Guideline principles are broad and intended at the highest level to ensure that statisticians behave professionally to advance knowledge while avoiding harm. Good statistical practice consists of using transparent assumptions and appropriate methods to produce reproducible results with valid interpretations.
The pharmaceutical industry reputation has suffered in recent years with a resultant diminished belief in data conclusions from clinical research. Embracing a robust set of ethical guidelines relating to data, its analysis and interpretations based on it, can bolster confidence in ethical research practices through a clear articulation of practices and behaviors that we can emulate.
Some key elements of the Guidelines will resonate with statisticians in the biopharmaceutical space:
- Use methods and data that are relevant and appropriate, without favoritism, in a manner to produce valid, interpretable, and reproducible results
- Be clear about limitations of data and methods; describe analytical assumptions, including missing data mechanisms and imputations.
- Support valid inferences, transparency, and good science.
- Respect privacy and confidentiality requirements of data collection, release, and sharing, e.g. stay within the limits of consents obtained.
- Understand and adhere to relevant rules, approvals, and guidelines protecting the rights and safety of subjects.
- Use adequate numbers of subjects to enable robust analytical conclusions and avoid excessive risk.
- Ensure that the discussion and reporting of statistical design and analysis is consistent with these Guidelines
- Avoid compromising scientific validity for expediency
- Strive to promote transparency in design, execution, and reporting or presenting of analyses.
- Promote sharing of data and methods as appropriate to facilitate replicate analyses, metadata studies, and other research by qualified investigators.
I would encourage you to read the
Guidelines, use them to support you in your daily work, and reach out to the Committee with examples that we can use as case studies. Case studies can be useful for both new and ongoing professional development, to strengthen understanding, and highlight the relevance, of these Guidelines in helping our profession.
Marcia Levenstein, Pfizer