2020
DOI: 10.1002/cncr.33369
|View full text |Cite
|
Sign up to set email alerts
|

Assessing risk factors with information beyond P value thresholds: Statistical significance does not equal clinical importance

Abstract: I thank the reviewers for their thoughtful comments as well as my colleagues Dr. Bonnie Ky (for many thoughtful discussions about risk factors for cardiotoxicity in patients with cancer), Dr. Jeffrey Min (for suggesting coefplot in Stata), and Ms. Jiali Yan (for suggesting SGPLOT in SAS).

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1

Citation Types

0
3
0

Year Published

2022
2022
2024
2024

Publication Types

Select...
5

Relationship

2
3

Authors

Journals

citations
Cited by 5 publications
(3 citation statements)
references
References 15 publications
0
3
0
Order By: Relevance
“…Thus, while impaired MFE was reasonably common (24.6% of the cohort), specific conditions were relatively rare and the power to test meaningful hypotheses presumably small. 27 Given this limitation, we focus on describing effect sizes, rather than P values, with the goal of informing future studies. For variables used in the earlier definition of impaired MFE, the OR for all but tobacco use was at least 1.8, albeit with wide CIs (Table 5 ).…”
Section: Resultsmentioning
confidence: 99%
“…Thus, while impaired MFE was reasonably common (24.6% of the cohort), specific conditions were relatively rare and the power to test meaningful hypotheses presumably small. 27 Given this limitation, we focus on describing effect sizes, rather than P values, with the goal of informing future studies. For variables used in the earlier definition of impaired MFE, the OR for all but tobacco use was at least 1.8, albeit with wide CIs (Table 5 ).…”
Section: Resultsmentioning
confidence: 99%
“…Hypothesis tests were based on Wald tests using a small-sample Satterthwaite correction to the degrees of freedom (Luke, 2017) and were two-sided without adjustment for multiple comparisons. Because the number of animals, and thus the statistical power to detect a specific effect, differed across interventions, ranking the efficacy of the interventions based on p-values would be misleading (Putt, 2021). Notably p-values lack context without an a priori power calculation based on assumptions about effect size and standard deviation.…”
Section: Statistical Analysesmentioning
confidence: 99%
“… 4 Again, an exhaustive review is intractable. For some examples, see Amrhein, Greenland, and McShane (2019a), Bernard (2019), Bijak (2019), Bresee (2019), Curran-Everett (2019), Davidson (2019), De Koning and Noordhof (2019), Dirnagl (2019), Harrington et al (2019), Harvey and Brinkhof (2019), Hayat et al (2019), Lowe (2019), Marshall (2019), McShane et al (2019), Morken (2019), Nguyen, Rivadeneira, and Civitelli (2019), Pickler (2019), O’Connor (2019), Parsons et al (2019), Staggs (2019), Carlsson and Gönen (2020), Charlesworth and Pandit (2020), Curran-Everett (2020), Johnson et al (2020), Knottnerus and Tugwell (2020), Marshall and Hughes (2020), Maula and Stam (2020), Michel, Murphy, and Motulsky (2020), Price, Bethune, and Massey (2020), Santibáñez, García-Rivero, and Barreiro (2020), Van Witteloostuijn (2020), Heckelei et al (2021), Imbens (2021), Putt (2021), Robinson and Haviland (2021), Tijssen (2021), Amrhein and Greenland (2022), Butler (2022), Elkins et al (2022), Filippini and Vinceti (2022), Greenland, Mansournia, and Joffe (2022), Bonovas and Piovani (2023), Fingerhut (2023), Hassler (2023), Montero, Hedeland, and Balgoma (2023), and Verykouki and Nakas (2023).…”
mentioning
confidence: 99%