ABSTRACT
 Traditional epidemiological assessments, which mainly focused on evaluating the statistical association between two major componentsthe exposure and outcomehave recently evolved to ascertain the inbetween process, which can explain the underlying causal pathway. Mediation analysis has emerged as a compelling method to disentangle the complex nature of these pathways. The statistical method of mediation analysis has evolved from simple regression analysis to causal mediation analysis, and each amendment refined the underlying mathematical theory and required assumptions. This short guide will introduce the basic statistical framework and assumptions of both traditional and modern mediation analyses, providing examples conducted with realworld data.

Keywords: Mediation analysis, Epidemiology, Humans, Logic, Probability
INTRODUCTION
 In the early days, traditional analytic epidemiological methods mainly focused on the statistical association between two major variables: the exposure (E) and the outcome (Y). However, methods have evolved to explore the “black box” between the E and the Y by investigating the mechanism underlying the association and various pathways. In the same context, the mechanism has also been visualized as being near the center of “Chinese boxes,” or a set of nested boxes. The “black box” is presumed to contain factors, both above and below the level of the individual—the factors above the individual may contain items such as interpersonal dynamics and socioeconomic status, including items related to ethnicity and politics, whereas the factors below the individual level comprise genes, proteins, cells, and organ systems [1].
 Mediation analysis was developed to assess this “black box,” and psychologists and social scientists have utilized this framework particularly frequently. Mediation analysis can explore and evaluate biological or social mechanisms, thereby elucidating unknown biological pathways and/or aiding in policymaking [2]. However, because of advances in methodologies, including biostatistics, epidemiological research designs, and causal inference, traditional mediation analysis has evolved and been applied in various fields. In particular, the concept of mediation analysis has been especially appealing in social sciences and psychology. There are several overviews of these topics [36], and this study is a guide to the full literature.
TRADITIONAL REGRESSIONBASED MEDIATION ANALYSIS
 Mediation was initially hypothesized as a variable in the middle of a causal chain. Previously, most of the epidemiological reports focused on evaluating the simple association between E and Y as in Figure 1A. However, as in Figure 1B, it is shown that an E affects a mediator (M), which in turn affects an Y. The M fully mediates the effect from the E to the Y. However, situations were identified where the M does not fully mediate the effect of E on the Y, which led to the concept of partial mediation, as depicted in Figure 1C. As shown in Figure 1C, the effect of an E can be exerted directly on an Y (direct effect, path c’) or take a detour via a M (indirect effect, paths a and b). Initially, the criteria to be regarded as a M were that E should have a statistically significant association with M, and that M should also have a statistically significant association with Y. The initial criteria also included the condition that the mediation analysis could be performed only if there was a statistically significant association between E and Y; this significant relationship between E and Y should be no longer significant after controlling for the previous paths from E to M and M to Y. However, the latter two conditions were further criticized due to the existence of inconsistent and partial mediation, and were therefore omitted from the essential conditions needed for mediation analysis.
 In contrast to a moderator or confounder, a M is interpreted as involving a causal pathway between E and Y. A detailed definition of a M is provided in the work of Robins and Greenland [7]. The seminal work on this concept of a M or intervening variable was based on Judd and Kenny [8,9] and Baron and Kenny [10]’s article utilizing the regression method.
 In Judd and Kenny[8,9]’s difference of coefficients approach, mediation analysis can be conceptualized as utilizing two regressions, as follows. First, we run a simple regression analysis with E on Y without M to estimate path c’.
 Second, we carry out a multivariable regression with E and M to predict Y.
 In this case, as the coefficient B reflects the total effect (TE), the direct effect from the E to Y c’ shown in Figure 1C, corresponds to B_{1} in equation 2. The difference method calculates the indirect effect by subtracting the direct effect (c’) from the TE, as follows:
 This is a simple and widely used approach to screen for the possible presence of a M. However, the logistic regression method has been criticized for lacking a causal interpretation. The difference method has been used to check for mediation, but nonsignificant findings using this method do not exclude the chance of possible mediation [11].
 The other approach is the product method, which was introduced by Sobel and used by Baron and Kenny [10]. In this method, again, a multivariable regression is conducted with E and M to predict Y.
 However, the next step is to regress M on X and can be written as
 In equation 3, B reflects path a in Figure 1C, and B_{2} in equation 2 reflects b in Figure 1C. The coefficient of the indirect effect, B_{indirect}, is calculated by multiplying the 2 coefficients, B_{2} and B.
 Generally, when there is no interaction between an E and a M, these two methods coincide, except for logistic regression. In particular, for rare Ys (approximately under 10%) with no confounding factors, these 2 estimates will, from a practical standpoint, reflect the natural indirect effect (NIE), which will be discussed in the causal mediation section. The difference method is beneficial because there is no restriction of the M distribution; it can be continuous or categorical (including binary). In contrast, the product method requires a linear model to be applied for the M [11]. In situations with common Ys, especially when they are binary, a loglinear regression model instead of logistic regression is recommended [12].
 To calculate the confidence interval (CI) of the indirect effect, 2 approaches have been suggested. The first approach utilizes the Sobel test, which is based on the product of 2 normally distributed values of coefficients. In this case, an assumption should be made about the shape of the sampling distribution of the indirect effect. The second approach uses resampling methods, such as bootstrap testing, which does not require a prior assumption of the sampling distribution. Usually, the bootstrap method involves resampling at least 750 times, for which reason the default resampling setting is 1000 times in many macros (e.g., R and the PROCESS macro in SAS [13,14]).
EXAMPLE OF REGRESSIONBASED MEDIATION ANALYSIS
 Kim et al. [15] conducted a study to estimate the mediating effect of lifestyle factors on the association between social networks and metabolic syndrome, utilizing the baseline data of the communitybased Cardiovascular and Metabolic Diseases Etiology Research Center cohort. In total, 10 103 participants were recruited from 2013 to 2018, and their egocentric social network properties were measured using a social network card that was previously applied and standardized [16]. From the raw data of the social network cards, the authors extracted and calculated the size of the social network and the closeness of the social network, which were used as quantitative E variables. Measurements of blood pressure, the lipid profile, fasting glucose, and waist circumference were made in the initial cohort, and metabolic syndrome was defined based on the National Cholesterol Education Program Adult Treatment Panel III criteria as the presence of 3 or more criteria.
 As potential Ms, the authors tested 4 domains: physical inactiveness (3 categories: vigorous activities, moderate activities, and walking), alcohol consumption (binary variable: current drinker vs. nondrinker), cigarette smoking (binary variable: current smoker vs. nonsmoker), and depressive symptoms (continuous variable: range 063 by Beck Depressive InventoryII score).
 After conducting the multivariable logistic regression for the E (social network properties, continuous variables) and Y (metabolic syndrome, yes/no), mediation analysis was performed with the ‘mediation’ package developed by Imai et al. [17] in the R software [18]. The analysis was conducted in 3 steps: (1) producing a M model, (2) producing an Y model, and (3) conducting a mediation analysis and sensitivity analysis. In the M model, social network properties and other covariates were regressed to explain lifestyle factors. The metabolic syndrome variable was then regressed on social network properties, lifestyle factors, and other covariates. These two models were grouped with the “mediate” function, which was run to estimate the direct effect, indirect effect, and their 95% CI by a quasiBayesian Monte Carlo method, including 5000 simulations per estimate set.
 As there were 4 potential Ms, the authors applied each M and tested the indirect effect. They found that only physical activity significantly mediated the relationship between social network size and metabolic syndrome in both genders (men: effect size [ES]=5.2×10^{3}, p=0.024; women: ES=3.1×10^{3}, p<0.001) (Figure 2A)
INTRODUCING CAUSAL MEDIATION ANALYSIS
 After the rise of the counterfactual framework for modern causal inference, the traditional approach in mediation analyses was expanded and redeveloped to solve the previous limitations regarding nonlinearities and interactions, focusing on the decomposition of direct and indirect effects [19,20]. Among the major issues raised, assumptions related to confounding factors and the interaction between the E and the M were reflected and redeveloped in causal mediation analysis [7,21]. In the counterfactual concept, an individual is hypothetically compared under an E and in the absence of the E in identical situations, including time and surrounding conditions. If the potential Ys are different based on this comparison, the E is regarded as causal for the Y [22].
 In causal mediation analysis, 3 terms regarding the previous indirect and direct effects are suggested. The natural direct effect (NDE) and NIE can be interpreted in traditional mediation analysis. There would be a difference between the counterfactual Ys if an individual was exposed to 2 different counterfactual situations, where the M value would be random at the reference value of the E. In contrast, the controlled direct effect (CDE) is different regarding the mediation value used in the calculation since the M is set to a certain fixed level. If there is no interaction between E and M, then the CDE usually coincides with the NDE [4].
 For example, an analysis using the NDE would ask “how much would the Y (e.g., suicide rate) change if the E was set at e=1 versus e=0 (e.g., exercise program), but for each participant, the M (e.g., the Patient Health Questionnaire [PHQ]9) was kept at the level it would have been in the absence of the E (i.e., the mean depressive symptom score of the group that did not participate in the exercise program)?” An analysis using the CDE would ask, “how much would the Y (e.g., suicide rate) would change on average if the M was controlled at a certain level (e.g., PHQ9=5) uniformly in the population?” Likewise, an analysis using the NIE would answer the question, “how much would the Y (e.g., suicide rate) would change on average if the E was controlled at the level it would be with the E present (e.g., with everyone participating in the exercise program), but with the M (e.g., PHQ9 change) changed from the level it would be with the E at the reference level (e.g., the usual rate of people in the exercise program) to the level it would be if the E is present?” In sum, the TE would correspond to the question, “how much would the Y (e.g., suicide rate) change overall with a change in the E from the reference value to the present?” This implies that the sum of the NDE and NIE equals the TE. Generally, the CDE has received more interest for policy evaluations, whereas the NIE and NDE have been used to elucidate the actions of various biological mechanisms.
 Similar to traditional mediation analysis, causal mediation analysis presumes the following temporal ordering: the E must precede the M measurement, and the Y measurement is performed after the M measurement. In addition, to interpret the mediation causally, 4 other assumptions related to confounding should be satisfied. First, all the known confounders should be controlled, and there should be no unmeasured confounding of the EY relationship (C_{1}) (Figure 3). If the E is randomized (e.g., in randomized clinical trials), this assumption will be met. Second, all the known confounders should be controlled, and there should be no unmeasured confounding of the MY relationship (C_{2}). In this case, it would not be enough to randomize only the E. Third, there should be no unmeasured confounding of the EM relationship, or all the known confounders should be controlled, which would be covered by E randomization. Lastly, there should be no confounding related to the MY relationship affected by the E, which means there is no arrow from E to C_{2} in Figure 3. As mentioned previously, randomizing the E (or treatment) is not enough to completely solve the confounding issue; randomizing E (which gives a probable even distribution of C_{1}) would not be sufficient to control the confounding, which can also occur between the M and Y, represented as C_{2}. In this case, conducting several sensitivity analyses would help, including situations with unmeasured confounding. Most importantly, it is strongly recommended to construct a directed acyclic graph depicting the central hypothesis before conducting a causal mediation analysis.
 In 2013, SAS (SAS Institute Inc., Cary, NC, USA) macros were used to perform a causal mediation analysis by Valeri and VanderWeele [2]. This initial macro dealt with binary forms of E, binary forms of Ms, and continuous Y variables. Additionally, in this macro, count variables could be applied as the Ys. A full description of this macro has been published elsewhere [4].
EXAMPLE OF CAUSAL MEDIATION ANALYSIS
 Lee et al. [23] performed a longitudinal analysis using data from 3347 participants aged 4064 years in the Korean Genome and Epidemiology Study, who were followed up for 16 years. As the E, socioeconomic status, including educational attainment and monthly household income, were queried at the index year and categorized into 2 groups. As the Y, sleep quality was queried with the Pittsburgh Sleep Quality Index at 5 time points (years 2, 6, 8, 10, and 12). As a M, depressive symptoms were measured using the Beck Depression Inventory at year 4. Sleep quality patterns were the Y variable. Using latent class growth modeling with SAS Proc traj syntax, a groupbased modeling approach was performed, and 5 subgroups were identified according to the pattern of sleep quality (“normalstable,” “moderatestable,” “poorstable,” “developing to poor,” and “severely poorstable”).
 Using SAS Proc causalmed syntax, the potential mediation of depressive symptoms on the association between socioeconomic factors and longitudinal sleep quality patterns was tested. Based on the maximum likelihood method, this SAS procedure estimates the effect of causal mediation and CIs from 1000 bootstrap replications [24]. Since this procedure permits a binary Y only, the original 5 sleep quality patterns were grouped into 2 categories, including a reference category (e.g., normalstable vs. moderatestable, or normalstable vs. severely poorstable). Percentages were calculated to explain the mediation and interaction effects, and the percentage of the TE after controlling the level of the M was also calculated [24].
 Overall, the associations between socioeconomic status variables and sleep patterns were not significant after full adjustment. However, depressive symptoms tended to fully mediate the associations between education/income variables and sleep quality patterns (e.g., for E=lower education vs. higher education, Y=developing to poor vs. normalstable, TE: odds ratio [OR], 1.55; 95% CI, 0.64 to 6.03; NDE: OR, 1.38; 95% CI, 0.58 to 5.09); NIE: OR, 1.12; 95% CI, 1.04 to 1.24) (Figure 2B).
CONCLUSION
 This paper reviewed the basic concepts of traditional mediation and causal mediation analysis with counterfactual approaches and provided examples in realworld settings.
 One issue to be aware of is that a statistically significant association regarding M in the mediation analysis (e.g., a statistically significant indirect effect) does not always confirm that M is an actual M. Using different causal models does not make it possible for researchers to prove a unique M unless it is theoretically plausible. Furthermore, mediation analysis itself cannot provide that an intervening variable is a true M by probabilistic inference, since we cannot verify the likelihood distribution of all other potential Ms and alternative causal models [25]. Therefore, it is essential to understand that researchers should interpret mediation analysis within the logic of theoretical inferences.
 Another issue lies in the measurement error for the M. According to a study conducted by le Cessie et al. [26], under the classical condition of a normally distributed M with nondifferential misclassification, the estimated mediated association tended toward the null. If the direct and indirect effects were the same, the estimates tended away from the null. However, when the M was multinomial, this pattern did not always exist. Correction methods, such as using a weighting coefficient and attenuating the regression coefficient B2 in equation 2, were also suggested by le Cessie et al. [26].
 Theoretical concepts and statistical application methods regarding mediation analysis are rapidly developing. As a result, further discussions on filling the gap between theoretical assumptions and practical analytical issues are required. It has been suggested that conceptualization and formalism may be obstacles for epidemiologists to apply these methods to actual analysis [27] and future directions should involve the development of more unified and simple methods that could be utilized by a broader base of users. However, because of its usefulness in elucidating complex mechanisms in population data, the rapid adoption of mediation analysis in future epidemiological studies is expected.
Ethics Statement
 As this review does not involve newly collected human data, institutional review board approval is not needed.
Notes

^{} The author has no conflicts of interest associated with the material presented in this paper.

^{} FUNDING
This research was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT (2020R1C1C1003502) and a faculty research grant of Yonsei University College of Medicine for 2019 (620190114).
ACKNOWLEDGEMENTS
None.
Notes

^{} AUTHOR CONTRIBUTIONS
All work was done by SJJ.
Figure. 1.A conceptual diagram of mediation analysis (A) traditional epidemiological assessment, (B) full mediation, and
(C) partial mediation.
Figure. 2.Brief conceptual diagrams of examples in this review. (A) Brief conceptual diagram by Kim et al. 2020 [15]. (B) Brief conceptual diagram by Lee et al. 2021 [23]. NDE, natural direct effect; OR, odds ratio; CI, confidence interval; NIE, natural indirect effect; TE, total effect. ^{*}p<0.05.
Figure. 3.Confounding assumptions in causal mediation analysis.
REFERENCES
 1. Weed DL. Beyond black box epidemiology. Am J Public Health 1998;88(1):1214ArticlePubMedPMC
 2. Valeri L, VanderWeele TJ. SAS macro for causal mediation analysis with survival data. Epidemiology 2015;26(2):e23e24ArticlePubMed
 3. VanderWeele TJ. Explanation in causal inference: methods for mediation and interaction. Oxford: Oxford University Press; 2015. p. 3245
 4. Valeri L, Vanderweele TJ. Mediation analysis allowing for exposuremediator interactions and causal interpretation: theoretical assumptions and implementation with SAS and SPSS macros. Psychol Methods 2013;18(2):137150ArticlePubMedPMC
 5. VanderWeele TJ. A unification of mediation and interaction: a 4way decomposition. Epidemiology 2014;25(5):749761ArticlePubMedPMC
 6. Hayes AF. Introduction to mediation, moderation, and conditional process analysis: a regressionbased approach. New York: Guilford Press; 2013. p. 3216
 7. Robins JM, Greenland S. Identifiability and exchangeability for direct and indirect effects. Epidemiology 1992;3(2):143155ArticlePubMed
 8. Judd CM, Kenny DA. Estimating the effects of social interventions. Cambridge: Cambridge University Press; 1981. p. 103131
 9. Judd CM, Kenny DA. Process analysis: estimating mediation in treatment evaluations. Eval Rev 1981;5(5):602619ArticlePDF
 10. Baron RM, Kenny DA. The moderatormediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. J Pers Soc Psychol 1986;51(6):11731182ArticlePubMed
 11. Jiang Z, VanderWeele TJ. When is the difference method conservative for assessing mediation? Am J Epidemiol 2015;182(2):105108ArticlePubMedPMC
 12. Lange T, Vansteelandt S, Bekaert M. A simple unified approach for estimating natural direct and indirect effects. Am J Epidemiol 2012;176(3):190195ArticlePubMed
 13. Hayes AF. Introduction to mediation, moderation, and conditional process analysis: a regressionbased approach. New York: Guilford Press; 2013. p. 551632
 14. Preacher KJ, Hayes AF. Asymptotic and resampling strategies for assessing and comparing indirect effects in multiple mediator models. Behav Res Methods 2008;40(3):879891ArticlePubMed
 15. Kim K, Jung SJ, Baek JM, Yim HW, Jeong H, Kim DJ, et al. Associations between social network properties and metabolic syndrome and the mediating effect of physical activity: findings from the Cardiovascular and Metabolic Diseases Etiology Research Center (CMERC) Cohort. BMJ Open Diabetes Res Care 2020;8(1):e001272ArticlePubMedPMC
 16. Herz A, Petermann S. Beyond interviewer effects in the standardized measurement of egocentric networks. Soc Networks 2017;50: 7082Article
 17. Imai K, Keele L, Tingley D, Yamamoto T. Causal mediation analysis using R. In: Vinod H, editor. Advances in social science research using R. New York: Springer; 2010. p. 129154
 18. Tingley D, Yamamoto T, Hirose K, Keele L, Imai K. Mediation: R package for causal mediation analysis. J Stat Softw 2014;59(5):38
 19. VanderWeele TJ. Simple relations between principal stratification and direct and indirect effects. Stat Probab Lett 2008;78(17):29572962Article
 20. Pearl J. Direct and indirect effects. 2001 [cited 2021 Feb 1]. Available from: https://arxiv.org/ftp/arxiv/papers/1301/1301.2300.pdfArticle
 21. VanderWeele TJ. Mediation analysis: a practitioner’s guide. Annu Rev Public Health 2016;37: 1732ArticlePubMed
 22. Berzuini C, Dawid P, Bernardinell L. Causality: statistical perspectives and applications. Chichester: John Wiley & Sons; 2012. p. 611
 23. Lee GB, Kim HC, Jeon YJ, Jung SJ. Association between socioeconomic status and longitudinal sleep quality patterns mediated by depressive symptoms. Sleep 2021;zsab044ArticlePubMedPMC
 24. Yung YF, Lamm M, Zhang W; SAS Institute Inc. Causal mediation analysis with the CAUSALMED procedure. 2018 [cited 2021 Feb 1]. Available from: https://www.sas.com/content/dam/SAS/support/en/sasglobalforumproceedings/2018/19912018.pdf
 25. Fiedler K, Schott M, Meiser T. What mediation analysis can (not) do. J Exp Soc Psychol 2011;47(6):12311236Article
 26. le Cessie S, Debeij J, Rosendaal FR, Cannegieter SC, Vandenbroucke JP. Quantification of bias in direct effects estimates due to different types of measurement error in the mediator. Epidemiology 2012;23(4):551560ArticlePubMed
 27. Lange T, Vansteelandt S, Bekaert M. A simple unified approach for estimating natural direct and indirect effects. Am J Epidemiol 2012;176(3):190195ArticlePubMed
Citations
Citations to this article as recorded by
 Residential greenspace and blood lipids in an essential hypertension population: Mediation through PM2.5 and chemical constituents
Ruoyi Lei, Ling Zhang, Xin Liu, Ce Liu, Ya Xiao, Baode Xue, Zengwu Wang, Jihong Hu, Zhoupeng Ren, Bin Luo
Environmental Research.2024; 240: 117418. CrossRef  The mediation roles of intermuscular fat and inflammation in muscle mitochondrial associations with cognition and mobility
Qu Tian, Philip R. Lee, Qi Yang, Anne Z. Moore, Bennett A. Landman, Susan M. Resnick, Luigi Ferrucci
Journal of Cachexia, Sarcopenia and Muscle.2024; 15(1): 138. CrossRef  Mediators of the Association Between Socioeconomic Status and Survival After OutofHospital Cardiac Arrest: A Systematic Review
Nicholas Grubic, Braeden Hill, Katherine S. Allan, Katerina Maximova, Hailey R. Banack, Marina del Rios, Amer M. Johri
Canadian Journal of Cardiology.2024; 40(6): 1088. CrossRef  Airway inflammation accelerates pulmonary exacerbations in cystic fibrosis
Theodore G. Liou, Natalia Argel, Fadi Asfour, Perry S. Brown, Barbara A. Chatfield, David R. Cox, Cori L. Daines, Dixie Durham, Jessica A. Francis, Barbara Glover, My Helms, Theresa Heynekamp, John R. Hoidal, Judy L. Jensen, Christiana Kartsonaki, Ruth Ke
iScience.2024; 27(3): 108835. CrossRef  Association Between Improved Serum Fatty Acid Profiles and Cognitive Function During a Dietary Intervention Trial in RelapsingRemitting Multiple Sclerosis
Solange M. Saxby, Carlyn Haas, Farnoosh Shemirani, Tyler J. Titcomb, Patrick Ten Eyck, Linda M. Rubenstein, Karin F. Hoth, Linda G. Snetselaar, Terry L. Wahls
International Journal of MS Care.2024; 26(2): 61. CrossRef  Effects of Covid19related anxiety on overeating and weight gain in a diverse college sample
Stephanie Guzman, Robert D. Melara
Journal of American College Health.2024; : 1. CrossRef  Childhood undernutrition mediates the relationship between open defecation with anemia among Ethiopian children: a nationally representative crosssectional study
Biniyam Sahiledengle, Pammla Petrucka, Fikreab Desta, Yordanos Sintayehu, Telila Mesfin, Lillian Mwanri
BMC Public Health.2024;[Epub] CrossRef  DietInduced Changes in Functional Disability among People with Multiple Sclerosis: A Secondary Pooled Analysis of Two Randomized Controlled Pilot Trials
Allison R. Groux, Elizabeth S. Walker, Farnoosh Shemirani, Jennifer E. Lee, Amanda K. Irish, Linda M. Rubenstein, Linda G. Snetselaar, Warren G. Darling, Terry L. Wahls, Tyler J. Titcomb
Sclerosis.2024; 2(3): 156. CrossRef  School Violence Exposure as an Adverse Childhood Experience: Protocol for a Nationwide Study of Secondary Public Schools (Preprint)
Sonali Rajan, Navjot Buttar, Zahra Ladhani, Jennifer Caruso, John Allegrante, Charles Branas
JMIR Research Protocols.2024;[Epub] CrossRef  Perseverance reduces whereas impulsivity increases the risk of reoffending
Marie Joséphine Hamatschek, Melanie S. Richter, KlausPeter Dahle
Forensische Psychiatrie, Psychologie, Kriminologie.2024;[Epub] CrossRef  Mixed heavy metals exposure affects the renal function mediated by 8OHG: A crosssectional study in rural residents of China
Xiaoyu Tian, Xiaobing Shan, Li Ma, Chenyang Zhang, Mei Wang, Jie Zheng, Ruoyi Lei, Li He, Jun Yan, Xun Li, Yanjun Bai, Keqin Hu, Sheng Li, Jingping Niu, Bin Luo
Environmental Pollution.2023; 317: 120727. CrossRef  Association between the overall burden of comorbidity and Ct values among the older patients with Omicron infection: Mediated by inflammation
Meixia Wang, Hongfei Mi, Na Li, Qingfeng Shi, Wei Sun, Tingjuan He, Jiabing Lin, Wenting Jin, Xiaodong Gao, Bijie Hu, Chenghao Su, Jue Pan
Frontiers in Immunology.2023;[Epub] CrossRef  Preferences of Young Polish Renters: Findings from the Mediation Analysis
Janusz Sobieraj, Marek Bryx, Dominik Metelski
Buildings.2023; 13(4): 920. CrossRef  Association of serum homocysteine, folate, and vitamin B12 and mood following the Swank and Wahls elimination dietary interventions in relapsingremitting multiple sclerosis: Secondary analysis of the WAVES trial
Farnoosh Shemirani, Tyler J. Titcomb, Solange M. Saxby, Patrick Ten Eyck, Linda M. Rubenstein, Karin F. Hoth, Linda G. Snetselaar, Terry L. Wahls
Multiple Sclerosis and Related Disorders.2023; 75: 104743. CrossRef  Mediation Analysis of Maternal Smoking, Gestational Age, and Birth Weight on the Texas–Mexico Border
Chinodebem Ogbutor, Stephanie M. Mishaw, Zuber D. Mulla
Southern Medical Journal.2023; 116(6): 478. CrossRef  Individual exposure of ambient particulate matters and eosinophilic chronic rhinosinusitis with nasal Polyps: DoseResponse, mediation effects and recurrence prediction
Jiajia Wang, Shen Shen, Bing Yan, Yunsheng He, Guoji Zhang, Chunguang Shan, Qintai Yang, Long Qin, Zhijian Duan, Luyun Jiang, Xin Wang, Xin Wei, Chengshuo Wang, Rui Chen, Luo Zhang
Environment International.2023; 177: 108031. CrossRef  The association between lead exposure and crime: A systematic review
Maria Jose Talayero, C. Rebecca Robbins, Emily R. Smith, Carlos SantosBurgoa, Naveen Puttaswamy
PLOS Global Public Health.2023; 3(8): e0002177. CrossRef  Mediation analysis of mental health characteristics linking social needs to life satisfaction among immigrants
David Adzrago, Faustine Williams
SSM  Population Health.2023; 24: 101522. CrossRef  Maternal anemia and baby birth size mediate the association between short birth interval and underfive undernutrition in Ethiopia: a generalized structural equation modeling approach
Desalegn Markos Shifti, Catherine Chojenta, Elizabeth G. Holliday, Deborah Loxton
BMC Pediatrics.2022;[Epub] CrossRef  Association of Elevated Maternal Psychological Distress, Altered Fetal Brain, and Offspring Cognitive and SocialEmotional Outcomes at 18 Months
Yao Wu, Kristina M. Espinosa, Scott D. Barnett, Anushree Kapse, Jessica Lynn Quistorff, Catherine Lopez, Nickie Andescavage, Subechhya Pradhan, YuanChiao Lu, Kushal Kapse, Diedtra Henderson, Gilbert Vezina, David Wessel, Adré J. du Plessis, Catherine Lim
JAMA Network Open.2022; 5(4): e229244. CrossRef