Methods for Evaluating Causality in Observational Studies

Validation of causal inference data using DirectLiNGAM in an environmental small-scale model and calculation settings

The development of data science has been needed in environmental fields such as marine, weather, and soil data. In general, the datasets are large in some cases, but they are often small because they contain observation data that the analyses themselves are limited. In such a case, the data are statistically evaluated by increasing or decreasing the levels of factors using differential analysis, resulting in the essential factors are estimated. However, there is no consistent approach to the means of assessing strong associations as a group between factors. Causal inference method has the possibility to output effective results for small data, and the results are expected to provide important information for understanding the potential highly association between factors, not necessarily the inference with big data. Here, we describe essential checkpoints and settings for the calculation by a direct method for learning a linear non-Gaussian structural equation model (DirectLiNGAM) and validation methods for the calculation results by using DirectLiNGAM with small-scale model data as an additional discussion of DirectLiNGAM portion of the related research article. Thus, this study provides the statistical validation methods for the association networks, treatments, and interventions for structural inference as a group of essential factors.•Causal inference with DirectLiNGAM•Validation of correlation coefficient and feature importance•Validation using causal effect object and propensity scores.

404-error "Disease not found": Unleashing the translational potential of -omics approaches beyond traditional disease classification in heart failure research

The emergence of personalized medicine, facilitated by the progress in -omics technologies, has initiated a new era in medical diagnostics and treatment. This review examines the potential of -omics approaches in heart failure, a condition that has not yet fully capitalized on personalized strategies compared to other medical fields like cancer therapy. Here, we argue that integrating multi-omics technology with systems medicine approaches could fundamentally transform heart failure management, moving away from the traditional paradigm of 'one size fits all'. Our review examines how omics can enhance understanding of heart failure's molecular foundations and contribute to a more comprehensive disease classification. We draw attention to the current state of medical practice that only relies on clinical evidence and a number of standard laboratory tests. At the same time, we propose a shift towards a universal approach that uses quantitative data from multi-omics to unravel complex molecular interactions. The discussion centres around the potential of the transition as a means to enhance individual risk assessment and emphasizes management within clinical settings. While the use of omics in cardiovascular research is not recent, many past studies have focused only on a single omics approach. In order to achieve a better understanding of disease mechanisms, we explore more holistic approaches using genomics, transcriptomics, epigenomics, and proteomics. This review concludes with a call to action to adopt multi-omics in clinical trials and practice to pave the way for more personalized disease management and more effective heart failure interventions.

Dental caries and developmental defects of enamel in cerebral palsy: A meta-analysis

OBJECTIVE

This systematic review and meta-analysis aimed to compare the occurrence of dental caries and developmental defects of enamel (DDE) in individuals with and without cerebral palsy (CP).

MATERIALS AND METHODS

We conducted searches across five databases and the grey literature. Data were organized using EndNote 20. Reporting followed the MOOSE checklist. A random-effects model meta-analyses were conducted using RStudio, presenting results as mean difference (MD), odds ratio (OR), and 95% confidence interval (CI). The risk of bias of studies was analyzed using the Newcastle-Ottawa Scale, and the certainty of evidence was assessed using GRADE.

RESULTS

Among 1336 identified records, 25 studies involving 59,997 participants (mean age: 11.1 years) were included. Data of 12 were pooled into meta-analyses. No significant differences were found between CP and non-CP individuals across indices: DMFT (k = 7) (MD = 0.31; 95% CI [-0.42-1.05]), dmft (k = 4) (MD = 0.31; 95% CI [-0.50-1.14]), DMFS (k = 2) (MD = -0.61; 95% CI [-20.56-19.33]), dmfs (k = 3) (MD = 0.54; 95% CI [-1.09-2.17]), and DDE (k = 3) (OR = 0.80, 95% CI [0.09-7.31]). The certainty of evidence was very low.

CONCLUSION

Individuals with CP do not appear to differ significantly from those without CP in terms of dental caries experience and DDE.

Mediation Analysis in Medical Research

BACKGROUND

Mediation analysis addresses the question of the mechanisms by which an exposure causes an outcome. This article is intended to convey basic knowledge of statistical mediation analysis.

METHODS

Selected articles and examples are used to explain the principle of mediation analysis.

RESULTS

The goal of mediation analysis is to express an overall exposure effect as a combination of an indirect and a direct effect. For example, it might be of interest whether the increased risk of diabetes (outcome) due to obesity (exposure) is mediated by insulin resistance (indirect effect), and, if so, how much of a direct effect remains. In this example, insulin resistance is a potential mediator of the effect of obesity on the risk of diabetes. In general, for a mediation analysis to be valid, more confounders must be taken into account than in the estimation of the overall effect size. A regression-based approach can be used to ensure the consideration of all relevant confounders in a mediation analysis.

CONCLUSION

By decomposing the overall exposure effect into indirect and direct components, a mediation analysis can reveal not just whether an exposure causes an outcome, but also how. For a mediation analysis to be valid, however, multiple assumptions must be satisfied that cannot easily be checked, potentially compromising such analyses as compared to the estimation of an overall effect.

Social integration, solidarity, and psychological health of internally displaced persons in Cameroon: Exploring the role of community satisfaction

Background/Introduction

Conflict-induced displacement continue to be a pressing public policy concern due to the adverse effects on the well-being of internally displaced persons (IDPs). Understanding the integration process of these individuals, particularly the role of solidarity and social integration in fostering a sense of belonging and improving their community experiences and psychological well-being, remains unclear. This study investigates the role of community satisfaction in the relationship among social integration, solidarity, and psychological health of IDPs in Cameroon.

Methods

Cross-sectional data from 428 displaced individuals in the Ntui subdivision, Cameroon, were analyzed using structural equation modeling (SEM). This technique tested four latent variables: social integration, solidarity, community satisfaction, and psychological health.

Results

The results show a direct association between solidarity and both community satisfaction (β = 0.282; p < 0.001) and psychological health (β = 0.137; p < 0.01). Community satisfaction also has a direct relationship with the psychological health of the IDPs (β = 0.292; p < 0.001). An indirect effect of solidarity on psychological health through community satisfaction was observed (β = 0.084 (SE = 0.025, CI = [0.047, 0.129], p < 0.001)). However, no mediational effect of community satisfaction was found on the association between social integration and psychological health. The model accounted for 7% of the variance in community satisfaction (R2 = 0.07) and 20% of the variance in psychological health (R2 = 0.20). Conclusion: The findings indicate a strong sense of solidarity among the displaced persons in the studied communities in Cameroon, which promotes their psychological health via community satisfaction. As such, interventions should aim to foster solidarity among displaced persons to enhance their community experiences and psychological health.

Causal relationship between osteoarthritis with atrial fibrillation and coronary atherosclerosis: a bidirectional Mendelian randomization study of European ancestry

Background

Osteoarthritis (OA) is a degenerative disease with high prevalence. Some observational studies have shown that patients with osteoarthritis often have co-existing cardiovascular diseases (CVD) such as atrial fibrillation (AF) and coronary atherosclerosis (CA). However, there is still a lack of stronger evidence confirming the association between osteoarthritis and cardiovascular disease. In this study, we used a bidirectional two-sample Mendelian randomization study to investigate the relationship between OA with AF and CA.

Methods

OA data from the UK Biobank and arcOGEN (Arthritis Research UK Osteoarthritis Genetics, a study that aimed to find genetic determinants of osteoarthritis and elucidate the genetic architecture of the disease) integration were selected for the study (n = 417,596), AF data were obtained from six studies (n = 1,030,836), and coronary atherosclerosis data were derived from the FinnGen (n = 218,792). MR analysis was performed primarily using the Inverse variance weighted (IVW) method, with MR Egger, weighted median, simple mode, weighted mode as supplements, sensitivity analysis was performed using Cochran Q statistic, and leave-one-out analysis.

Results

We found that OA and AF were positively associated [IVW: OR (95% CI): 1.11 (1.04, 1.19), P = 0.002], while OA and CA were negatively associated [IVW: OR (95% CI): 0.88 (0.79, 0.98), P = 0.02]. In the reverse MR analysis, no effect of AF on OA was found [IVW: OR (95% CI): 1.00 (0.97, 1.03), P = 0.84], meanwhile, CA and OA were found to be associated negatively [IVW: OR (95% CI): 0.95 (0.92, 0.99), P = 0.01]. No violations of MR assumptions were found in the sensitivity analysis.

Conclusion

This research confirms that OA is a risk factor for AF, and there is a mutual protective factor between OA and CA. However, further studies are still necessary to elucidate the underlying mechanisms.

Machine learning models for 180-day mortality prediction of patients with advanced cancer using patient-reported symptom data

PURPOSE

The objective of the current study was to develop and test the performances of different ML algorithms which were trained using patient-reported symptom severity data to predict mortality within 180 days for patients with advanced cancer.

METHODS

We randomly selected 630 of 689 patients with advanced cancer at our institution who completed symptom PRO measures as part of routine care between 2009 and 2020. Using clinical, demographic, and PRO data, we trained and tested four ML algorithms: generalized regression with elastic net regularization (GLM), extreme gradient boosting (XGBoost) trees, support vector machines (SVM), and a single hidden layer neural network (NNET). We assessed the performance of algorithms individually as well as part of an unweighted voting ensemble on the hold-out testing sample. Performance was assessed using area under the receiver-operating characteristic curve (AUROC), sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV).

RESULTS

The starting cohort of 630 patients was randomly partitioned into training (n = 504) and testing (n = 126) samples. Of the four ML models, the XGBoost algorithm demonstrated the best performance for 180-day mortality prediction in testing data (AUROC = 0.69, sensitivity = 0.68, specificity = 0.62, PPV = 0.66, NPV = 0.64). Ensemble of all algorithms performed worst (AUROC = 0.65, sensitivity = 0.65, specificity = 0.62, PPV = 0.65, NPV = 0.62). Of individual PRO symptoms, shortness of breath emerged as the variable of highest impact on the XGBoost 180-mortality prediction (1-AUROC = 0.30).

CONCLUSION

Our findings support ML models driven by patient-reported symptom severity as accurate predictors of short-term mortality in patients with advanced cancer, highlighting the opportunity to integrate these models prospectively into future studies of goal-concordant care.

Elevation of cell-associated HIV-1 transcripts in CSF CD4+ T cells, despite effective antiretroviral therapy, is linked to brain injury

Antiretroviral therapy (ART) can attain prolonged undetectable HIV-1 in plasma and cerebrospinal fluid (CSF), but brain injury remains prevalent in people living with HIV-1 infection (PLHIV). We investigated cell-associated (CA)-HIV-1 RNA transcripts in cells in CSF and blood, using the highly sensitive Double-R assay, together with proton Magnetic Resonance Spectroscopy (¹H MRS) of major brain metabolites, in sixteen PLHIV. 14/16 CSF cell samples had quantifiable CA-HIV-1 RNA, at levels significantly higher than in their PBMCs (median 9,266 vs 185 copies /106 CD4+ T-cells; p<0.0001). In individual PLHIV, higher levels of HIV-1 transcripts in CSF cells were associated with greater brain injury in the frontal white matter (Std β=-0.73; p=0.007) and posterior cingulate (Std β=-0.61; p=0.03). 18-colour flow cytometry revealed that the CSF cells were 91% memory T-cells, equally CD4+ and CD8+ T-cells, but fewer B cells (0.4 %), and monocytes (3.1%). CXCR3⁺CD49d⁺integrin β7-, CCR5⁺CD4⁺ T-cells were highly enriched in CSF, compared with PBMC (p <0.001). However, CA-HIV-1 RNA could not be detected in 10/16 preparations of highly purified monocytes from PBMC, and was extremely low in the other six. Our data show that elevated HIV-1 transcripts in CSF cells were associated with brain injury, despite suppressive ART. The cellular source is most likely memory CD4⁺ T cells from blood, rather than trafficking monocytes. Future research should focus on inhibitors of this transcription to reduce local production of potentially neurotoxic and inflammatory viral products.

Evaluation of the Relationship between Left Coronary Artery Bifurcation Angle and Coronary Artery Disease: A Systematic Review

Recent studies have suggested a relationship between wide left coronary artery bifurcation (left anterior descending [LAD]-left circumflex [LCx]) angle and coronary artery disease (CAD). Current literature is multifaceted. Different studies have analysed this relationship using computational fluid dynamics, by considering CAD risk factors, and from simple causal-comparative and correlational perspectives. Hence, the purpose of this systematic review was to critically evaluate the current literature and determine whether there is sufficient evidence available to prove the relationship between LAD-LCx angle and CAD. Five electronic databases (ProQuest, Scopus, PubMed, CINAHL Plus with Full Text, and Emcare) were used to locate relevant texts, which were then screened according to predefined eligibility criteria. Thirteen eligible articles were selected for review. Current evidence suggests individuals with a wide LAD-LCx angle experience altered haemodynamics at the bifurcation site compared to those with narrower angles, which likely facilitates a predisposition to developing CAD. However, further research is required to determine causality regarding relationships between LAD-LCx angle and CAD risk factors. Insufficient valid evidence exists to support associations between LAD-LCx angle and degree of coronary stenosis, and future haemodynamic analyses should explore more accurate coronary artery modelling, as well as CAD progression in already stenosed bifurcations.

Osteoarthritis, coronary artery disease, and myocardial infarction: A mendelian randomization study

Background

Observational studies indicate that osteoarthritis (OA) and coronary artery disease (CAD), as well as myocardial infarction (MI), are often diagnosed as comorbid diseases. We performed a bidirectional Mendelian randomization (MR) study to demonstrate whether there is a causal relationship between OA, CAD, and MI.

Methods

We extracted single nucleotide polymorphisms (SNPs) related to OA in the Genetics of Osteoarthritis (GO) Consortium as instrumental variables to assess whether OA is associated with CAD and MI in the CARDIoGRAMplusC4D 1,000 Genomes genome-wide association study (GWAS). In the reverse MR, we used CAD-associated and MI-associated SNPs to the GWAS of OA to analyze their causality. These GWASs included 766,690 individuals of OA, 184,305 individuals of CAD, and 166,065 individuals of MI. MR was conducted using several methods, including the inverse variance weighted (IVW) method, the weighted median method, the MR-Egger method, and the MR-Pleiotropy RESidual Sum and Outlier (MR-PRESSO) method.

Results

The forward causal effect of OA on CAD and MI was not observed. In reverse analysis, no causal effect was discovered for CAD on the risk of OA. Notably, we observed a causal association between MI and total OA [IVW odds ratio (OR) = 0.95, 95% CI = 0.93, 0.98, P = 4E−04] and spine OA (IVW OR = 0.92, 95% CI = 0.88, 0.97, P = 0.001) but a null association between MI and knee OA, hip OA, hand OA, and thumb OA.

Conclusion

This MR study identifies a potentially protective effect of genetically predicted MI on total and spine OA risks.

Periodontal outcomes of children and adolescents with attention deficit hyperactivity disorder: a systematic review and meta-analysis

BACKGROUND

This systematic review and meta-analysis aimed to answer the following question: Are children and adolescents with attention deficit hyperactivity disorder (ADHD) more likely to have gingival or periodontal disease-related outcomes than their non-ADHD peers?

METHODS

Searches were conducted in the following databases: Embase, Scopus, Web of Science, and PubMed. Google Scholar and OpenGrey were also verified. Observational studies were included in which children and adolescents with ADHD were compared with their healthy peers in terms of gingival and/or periodontal endpoints. Bias appraisal was performed using the Joann Briggs tool for case-control and cross-sectional studies. Meta-analysis was performed using R language. Results are reported as mean difference (MD) and odds ratio (OR). Statistical analyses were performed in RStudio.

RESULTS

A total of 149 records were identified in the searches. Seven studies were included. The meta-analysis showed that children and adolescents with ADHD had a higher mean gingival bleeding index (percentage) than their non-ADHD peers (MD = 11.25; CI = 0.08-22.41; I² = 73%). There was no difference between groups for plaque index (MD = 4.87; CI = - 2.56 to 12.30; I² = 63%) and gingivitis (OR = 1.42; CI = 0.22-9.21; I² = 76%). Regarding the assessment of risk of bias, the major issue found in the articles was the absence of analyses for the control of confounding factors.

CONCLUSION

Children and adolescents with ADHD had more gingival bleeding than their non-ADHD peers, but no difference regarding plaque or gingivitis was detected between groups.

CLINICAL REGISTRATION

CRD42021258404.

COVID-19 Lockdown in Israel: The Environmental Effect on Ultrafine Particle Content in the Airway

Inhaled ultrafine particle (UFP) content in exhaled breath condensate (EBC) was observed as an airway inflammatory marker and an indicator of exposure to particulate matter (PM). The exceptional decline in air pollution during the COVID-19 lockdown was an opportunity to evaluate the effect of environmental changes on UFP airway content. We collected EBC samples from 30 healthy subjects during the first lockdown due to COVID-19 in Israel (March–April 2020) and compared them to EBC samples retrieved during April–June 2016 from 25 other healthy subjects (controls) living in the same northern Israeli district. All participants underwent EBC collection and blood sampling. Ambient air pollutant levels were collected from the Israeli Ministry of Environmental Protection’s online database. Data were acquired from the monitoring station closest to each subject’s home address, and means were calculated for a duration of 1 month preceding EBC collection. UFP contents were measured in the EBC and blood samples by means of the NanoSight LM20 system. There was a dramatic reduction in NO, NO₂, SO₂, and O₃ levels during lockdown compared to a similar period in 2016 (by 61%, 26%, 50%, and 45%, respectively). The specific NO₂ levels were 8.3 ppb for the lockdown group and 11.2 ppb for the controls (p = 0.01). The lockdown group had higher UFP concentrations in EBC and lower UFP concentrations in serum compared to controls (0.58 × 10⁸/mL and 4.3 × 10⁸/mL vs. 0.43 × 10⁸/mL and 6.7 × 10⁸/mL, p = 0.05 and p = 0.03, respectively). In this observational study, reduced levels of air pollution during the COVID-19 lockdown were reflected in increased levels of UFP airway contents. The suggested mechanism is that low airway inflammation levels during lockdown resulted in a decreased UFP translocation to serum. Further studies are needed to confirm this hypothesis.

Case Report: Plasticity in Central Sensory Finger Representation and Touch Perception After Microsurgical Reconstruction of Infraclavicular Brachial Plexus Injury

After brachial plexus injury (BPI), early microsurgery aims at facilitating reconnection of the severed peripheral nerves with their orphan muscles and sensory receptors and thereby reestablishing communication with the brain. In order to investigate this sensory recovery, here we combined functional magnetic resonance imaging (fMRI) and tactile psychophysics in a patient who suffered a sharp, incomplete amputation of the dominant hand at the axilla level. To determine somatosensory detection and discomfort thresholds as well as sensory accuracy for fingers of both the intact and affected hand, we used electrotactile stimulation in the framework of a mislocalization test. Additionally, tactile stimulation was performed in the MRI scanner in order to determine the cortical organization of the possibly affected primary somatosensory cortex. The patient was able to detect electrotactile stimulation in 4 of the 5 fingertips (D1, D2, D4, D5), and in the middle phalanx in D3 indicating some innervation. The detection and discomfort threshold were considerably higher at the affected side than at the intact side, with higher detection and discomfort thresholds for the affected side. The discrimination accuracy was rather low at the affected side, with stimulation of D1/D2/D3/D4/D5 eliciting most commonly a sensation at D4/D1/D3/D2/D5, respectively. The neuroimaging data showed a mediolateral succession from D2 to D5 to D1 to D4 (no activation was observed for D3). These results indicate a successful regrowth of the peripheral nerve fibers from the axilla to four fingertips. The data suggest that some of the fibers have switched location in the process and there is a beginning of cortical reorganization in the primary somatosensory cortex, possibly resulting from a re-education of the brain due to conflicting information (touch vs. vision).

Collider Bias in Observational Studies: Consequences for Medical Research–Part 30 of a Series on Evaluation of Scientific Publications

BACKGROUND

The findings of observational studies can be distorted by a number of factors. Socalled confounders are well known, but distortion by collider bias (CB) has received little attention in medical research to date. The goal of this article is to present the principle of CB, and measures that can be taken to avoid it, by way of a few illustrative examples.

METHODS

The findings of a selective review of the literature on CB are explained with illustrative examples.

RESULTS

The simplest case of a collider variable is one that is caused by at least two other variables. An example of CB is the observation that, among persons with diabetes, obesity is associated with lower mortality, even though it is associated with higher mortality in the general population. The false protective association between obesity and mortality arises from the restriction of the study population to persons with diabetes.

CONCLUSION

CB is a distortion that arises through restriction on or stratification by a collider variable, or through statistical adjustment for a collider variable in a regression model. CB can arise in many ways. The graphic representation of causal structures helps to identify potential sources of CB. It is important to distinguish confounders from colliders, as methods that serve to correct for confounding can themselves cause bias when applied to colliders. There is no generally applicable method for correcting CB.

Environmental assessment of interventions to restrain the impact of industrial pollution using a quasi-experimental design: limitations of the interventions and recommendations for public health policy

Background

In an industrial area, the asymmetry between the weights of the economic interests compared to the public-health needs can determine which interests are represented in decision-making processes. This might lead to partial interventions, whose impacts are not always evaluated. This study focuses on two interventions implemented in Taranto, Italy, a city hosting one of the largest steel plants in Europe. The first intervention deals with measures industrial plants must implement by law to reduce emissions during so called “wind days” in order to reduce PM₁₀ and benzo [a] pyrene concentrations. The second one is a warning to the population with recommendations to aerate indoor spaces from 12 pm to 6 pm, when pollutant concentrations are believed to be lower.

Methods

To analyse the impact of the first intervention, we analysed monthly PM₁₀ data in the period 2009–2016 from two monitoring stations and conducted an interrupted-time-series analysis. Coefficients of time-based covariates are estimated in the regression model. To minimise potential confounding, monthly concentrations of PM₁₀ in a neighbourhood 13 km away from the steel plant were used as a control series. To evaluate the second intervention, hourly concentrations of PM₁₀, SO₂ and polycyclic-aromatic-hydrocarbons (PAHs) were analysed.

Results

PM₁₀ concentrations in the intervention neighbourhood showed a peak just a few months before the introduction of the law. When compared to the control series, PM₁₀ concentrations were constantly higher throughout the entire study period. After the intervention, there was a reduction in the difference between the two time-series (− 25.6%). During “wind days” results suggested no reduction in concentrations of air pollutants from 12 pm to 18 pm.

Conclusion

Results of our study suggest revising the warning to the population. Furthermore, they evidence that in complex highly industrialised areas, air quality interventions cannot focus on only a single pollutant, but rather should consider the complex relationships between the different contaminants. Environmental interventions should be reviewed periodically, particularly when they have implications for social constraints. While the results of our study can be related only to the specific situation reported in the article, the methodology applied might be useful for the environmental management in industrial areas with similar features.

Health-Related Quality of Life and Frequency of Physical Activity in Spanish Students Aged 8-14

The study of health-related quality of life (HRQoL) in children and adolescents has important implications in terms of policy, education, and health. Data on the time spent in physical activity (PA) and in sedentary activities in this population are worrying. We aim to analyze possible differences in HRQoL and PA levels between sexes and age groups in Spanish students aged between 8 and 14 years, as well as to assess the relationship between HRQoL and the frequency of PA in this population. A total of 3197 participants (1610 boys and 1587 girls) from 8 to 14 years old were recruited. Mquality and Mapping Child Health Utility instrument (Chu9d) were used as HRQoL indicators. A medium positive association between PA and HRQoL concerning the Spanish school population was found. HRQoL was higher among students aged 8 to 12 than 13 to 14. Moreover, when children start secondary education, both sexes seem to lose the quality of life. Similarly, PA decreases among girls over the years, although it seems to increase among boys. Thus, PA levels and HRQoL are directly associated in Spanish schoolchildren aged between 8 and 14 years. However, this HRQoL decreases in children over the years. Practical implications include the need to support education and physical activity programs to improve HRQoL in children and adolescents.

Long-term impact of ambient air pollution on preterm birth in Xuzhou, China: a time series study

Accumulating evidence witnesses the negative influence of air pollution on human health, but the relationship between air pollution and premature babies has been inconsistent. In this study, the association between weekly average concentration of air pollutants and preterm birth (PTB) was conducted in Xuzhou, a heavy industry city, in China. We constructed a distributed lag non-linear model (DLNM), an ecological study, to access the associations between ambient air pollutants and PTB in this study. Totally, 5408 premature babies were included, and the weekly average levels of PM_2.5, PM₁₀, SO₂, NO₂, O₃, and CO were 61.24, 110.21, 22.55, 40.55, 104.45, and 1.04 mg/m³, respectively. We found that PM_2.5, PM₁₀, SO₂, and NO₂ significantly increased the risk of PTB, and the susceptibility windows of these contaminants were the second trimester and third trimester (from 12 to 29 weeks). Every 10 μg/m³ increase of PM_2.5, PM₁₀, SO₂, and NO₂, the greatest relative risk (RR) values and 95% confidence interval (CI) on PTB were 1.0075 [95% CI, 1.0019-1.0131], 1.0053 [95% CI, 1.0014-1.0092], 1.0203 [95% CI, 1.0030-1.0379], and 1.0170 [95% CI, 1.0052-1.0289] in lag 16th, 18th, 19th, and 20th gestational weeks, respectively. No significant influence of O₃ and CO were found on preterm birth. Subgroup analysis showed that the risk of premature delivery was higher for younger pregnant women and in warm season. This finding shows that prenatal exposure to ambient air pollutants is associated with preterm birth, and there existed an exposure window period.