Explore chapters and articles related to this topic
Relevant study design issues
Published in O. Ajetunmobi, Making Sense of Critical Appraisal, 2021
A ‘confounding factor’ is any factor associated with both the variable(s) of interest and the outcome of a study. It is often an unrecognized or even unmeasured factor that can adversely influence any observations we may make about the relationship between variable(s) and outcome in a study. The presence of confounding factors can therefore weaken the validity of the results obtained from a study.
Using Real-World Evidence to Transform Drug Development: Opportunities and Challenges
Published in Harry Yang, Binbing Yu, Real-World Evidence in Drug Development and Evaluation, 2021
As discussed by Skovlund et al. (2018), there are several remedies that can be used to control for potential confounding. One is to adjust for known confounding factors through statistical models. However, there exist so-called residual confounding caused by factors that are not measured or measurement error due to misclassification of the known confounding variables (Greenland 1996; Skovlund et al. 2018). Another potential solution is to use propensity score-based methods to match patients to different treatments according to key patient characteristics (see Chapters 4 and 7 for detailed discussion). But propensity scores also suffer the inability to balance characteristics that are not measured (Rubin 1997). The use of instrumental variables as a substitute for the actual treatment status is another alternative method, but it has its own challenges. For instance, it is difficult to find such valid instruments (Greenland 2000; Schneeweiss 2007; Burgess and Thompson 2011).
Meta-Analysis of Epidemiological Data, with a Focus on Individual Participant Data
Published in Christopher H. Schmid, Theo Stijnen, Ian R. White, Handbook of Meta-Analysis, 2020
Angela Wood, Stephen Kaptoge, Michael Sweeting, Clare Oliver-Williams
Exposure-disease associations estimated from observational data are subject to confounding by measured and unmeasured risk factors. Statistical adjustment of confounding factors measured a priori may help mitigate confounding bias to some extent. However, except for some well-recognized confounding factors in some circumstances (e.g., age, sex, smoking, diabetes, blood pressure, adiposity, and lipids for cardiovascular disease), the confounding factors available or considered by researchers would often vary by study (i.e., systematically missing).
Characteristic analysis of early gastric cancer after Helicobacter pylori eradication: a multicenter retrospective propensity score-matched study
Published in Annals of Medicine, 2023
Xinyuan Liu, Xinyu Wang, Tao Mao, Xiaoyan Yin, Zhi Wei, Jindong Fu, Jie Wu, Xiaoyu Li
Figure 1 is the flow diagram of this study. Firstly, the clinical characteristics of the Hp-eradicated and control groups before and after propensity-score matching are summarized in Table 1. A total of 81 patients were included in the Hp-eradicated group in this study, with an average age of 61.81 ± 9.2 years, and 105 patients were included in the control group, with an average age of 61.14 ± 8.9 years. Patients in the Hp-eradicated group had more clinical consultations (p = 0.003) and endoscopic examinations (p = 0.022) before propensity-score matching. No statistically significant differences were found in smoking history, drinking history, family history of GC, CCI, clinical symptoms, and duration of symptoms between the two groups. We performed propensity-score matched analysis to remove the influence of confounding factors and improve the reliability of results. After propensity-score matching, 62 patients from each group were included in the analysis. There were more patients in the Hp-eradicated group who were administered PPIs for >1 year compared to in the control group before and after propensity-score matching (p < 0.001).
The effects of different parity and delivery mode on wheezing disorders in the children—a retrospective cohort study in Fujian, China
Published in Journal of Asthma, 2022
Haiyan Gao, Chong Miao, Haibo Li, Meng Bai, Huijie Zhang, Zhengqin Wu, Wei Li, Wenjuan Liu, Libo Xu, Guanghua Liu, Yibing Zhu
SPSS software (version 26.0) was used for all calculations and analyses. Frequency (percentage) was used for categorical variables. Binary logistic regression analysis was used to analyze the categorical variables. Multivariate analyses were performed using a multivariable logistic regression analysis. The adjusted confounding factors were nationality, gender, age, dwelling environment, onset season, birth weight, jaundice history, complications, and feeding patterns. To adjust for the effects of confounding factors, an appropriate stratification analysis was performed. We showed different stratifications according to parity; parity and gender; and parity and birth weight. A cross-product term of the mode of childbirth and parity was constructed using the likelihood ratio test in the logistic regression model to assess multiplicative interaction. Statistical tests were interpreted at a two-sided significance level of 0.05. A total of 4.9% of cases had missing data for the delivery mode, and were excluded from the study. The missing values of other indicators that were much lower than 5% were not filled up.
Age-specific association between invasively measured central blood pressure and left ventricular mass index
Published in Clinical and Experimental Hypertension, 2021
Tae-Min Rhee, Hack-Lyoung Kim, Woo-Hyun Lim, Jae-Bin Seo, Sang-Hyun Kim, Joo-Hee Zo, Myung-A Kim
Several limitations should be discussed. First, as a cross-sectional study, we could not assess the causal relationship between CBP and LVMI, or evaluate clinical outcomes according to CBP indices. Secondly, possible confounding factors were adjusted as much as possible using multivariable analysis and propensity-score matching, but residual or unmeasured confounding effects could exist. Thirdly, the number of the study subjects, especially the younger age group, was relatively small. Although statistically significant results were presented, these results should be reproduced in future studies in a larger population. Finally, subjects in the younger age group were the main comparison group; however, a significant proportion of the group also had multiple cardiovascular risk factors or significant CAD. Therefore, it is limited to extrapolate the results in young and healthy populations at very low risk.