Multivariable models in orthopaedic research: a methodological review of covariate selection and causal relationships.

OBJECTIVES: The aim of this study was to evaluate the methods used for including or excluding covariates in a multivariable model and to find out how common is the Table 2 Fallacy in studies recently published in high-quality orthopaedic journals. METHODS: A systematic review was conducted in the MEDLINE database. We included all studies that presented the results of a multivariable model in a table and published in seven orthopaedic journals with the highest ranked impact factors in 2019. RESULTS: Table 2 Fallacy was found in 67% (129/193) of the evaluated studies in which a multivariable model was used. Only 16% (31/193) of all studies had included the variables based on causal inference. Furthermore, only three of these studies used causal diagrams to illustrate the causal inference. Altogether, 35% (67/193) of the studies included variables based on statistical methods. CONCLUSIONS: Confounder selection and the interpretation of the results of the multivariable model showed notable challenges in orthopaedic studies recently published in the top orthopaedic journals. Based on the results of our review, it seems that more education in statistics and increased knowledge is required to decrease the occurrence of these statistical issues in orthopaedic research.

Revisiting the Table 2 fallacy: A motivating example examining preeclampsia and preterm birth.

BACKGROUND: A "Table Fallacy," as coined by Westreich and Greenland, reports multiple adjusted effect estimates from a single model. This practice, which remains common in published literature, can be problematic when different types of effect estimates are presented together in a single table. The purpose of this paper is to quantitatively illustrate this potential for misinterpretation with an example estimating the effects of preeclampsia on preterm birth. METHODS: We analysed a retrospective population-based cohort of 2 963 888 singleton births in California between 2007 and 2012. We performed a modified Poisson regression to calculate the total effect of preeclampsia on the risk of PTB, adjusting for previous preterm birth. pregnancy alcohol abuse, maternal education, and maternal socio-demographic factors (Model 1). In subsequent models, we report the total effects of previous preterm birth, alcohol abuse, and education on the risk of PTB, comparing and contrasting the controlled direct effects, total effects, and confounded effect estimates, resulting from Model 1. RESULTS: The effect estimate for previous preterm birth (a controlled direct effect in Model 1) increased 10% when estimated as a total effect. The risk ratio for alcohol abuse, biased due to an uncontrolled confounder in Model 1, was reduced by 23% when adjusted for drug abuse. The risk ratio for maternal education, solely a predictor of the outcome, was essentially unchanged. CONCLUSIONS: Reporting multiple effect estimates from a single model may lead to misinterpretation and lack of reproducibility. This example highlights the need for careful consideration of the types of effects estimated in statistical models.

A scoping review of Table 2 fallacy in the oral health literature.

BACKGROUND: Coined by Westreich and Greenland in 2013, Table 2 fallacy refers to the practice of reporting estimates of the primary exposure and adjustment covariates derived from a single model on the same table. This study seeks to describe the extent to which Table 2 fallacy is present in the oral health literature and provide recommendations on presenting findings from multivariable-adjusted models and/or interpretation of adjustment covariate estimates that are not the primary exposure. METHODS: We conducted a scoping review in PubMed and Scopus of human observational studies published in 4 oral health journals (JDR-CTR, CDOE, JPHD, BMC Oral Health) starting in 2013 until the end of 2018. The resulting articles were exported into Excel and were either included or excluded for full-text review based on six criteria. After categorizing the articles, we exported and summarized the results in SAS. RESULTS: A total of 1358 articles were initially screened of which 937 articles were excluded based on title or abstract for being animal studies, systematic reviews or meta-analysis, prediction models or descriptive studies. The remaining 421 articles were eligible for full text reviewed of which, 189 (45%) committed Table 2 fallacy. The prevalence of table 2 fallacy appears high in the oral health literature. CONCLUSIONS: The problem of presenting multiple effect estimates derived from a single model in the same table is that it inadvertently encourages the reader to interpret all estimates the same way, often as total effects. Implications and recommendations are discussed.