CONTROLLING FOR MULTIPLICITY IN PURSUIT OF A PRO-BASED LABEL WHEN MULTIPLE PROS ARE ASSESSED
Author(s)
Cole JCCovance Market Access Services, Inc., San Diego, CA, USA
Presentation Documents
OBJECTIVES: The FDA's final Guidance for industry on patient reported outcome (PRO) use in support of labeling claims was issued in December, 2009. In their Guidance, the FDA noted that a study’s endpoint model must consider the hierarchy of multiple endpoints, including how PROs used for a label claim fit into this hierarchy. Whereas most studies implement a basic sequential gatekeeping process to articulate their hierarchy, this may place some potential labels at risk. Researchers should be knowledge of the various ways familywise error is influenced and how best to control for it with an informed multiplicity plan as part of their endpoint model. METHODS: Outcomes from previously published literature were examined for the influence of various familywise error issues and related multiplicity controls, including analytic issues, gatekeeping, and precision alpha control (vs. Bonferroni or Hochberg). RESULTS: In a study with one clinical and three PRO outcomes, A Bonferroni correction resulted in just one significant result. A gatekeep between primary and secondary outcomes resulted in two significant findings. Finally, when using either an adjustment for known-levels of correlation to adjust alpha (Tukey’s test of statistical certainty) or using a repeated measures ANOVA vs. change-score analysis, three of the outcomes were classified as significant. CONCLUSIONS: Researchers should understand the implications of their multiplicity control in order to make informed decisions about their analyses, organization of their endpoint model, and ultimately make the best plans to ensure their desired PRO-based label claims have the most accurate demonstration of their statistical probability.
Conference/Value in Health Info
2012-06, ISPOR 2012, Washington, D.C., USA
Value in Health, Vol. 15, No. 4 (June 2012)
Code
PRM46
Topic
Methodological & Statistical Research
Topic Subcategory
Confounding, Selection Bias Correction, Causal Inference
Disease
Multiple Diseases