Causal Association Between Lifestyle Behavior and Potential Risk Factors With Blepharoptosis

Abstract

Background: Blepharoptosis is among the most common disorders of eyelid malposition that may influence appearance and damage visual function, and both of these can have negative effects on the quality of life. Various types of blepharoptosis have been reported, including neurogenic, traumatic, congenital, mechanical, psychogenic, and myogenic, while about 15%–25% remain uncertain.

Method: To determine the targets for preventing blepharoptosis and its associated disease burden, we employed a two‐sample Mendelian randomization approach to explore the causal relationships between reported potential risk factors and lifestyle behaviors and blepharoptosis. We chose genetic variants (p < 5 × 10⁻⁸) related to prediabetic status, obesity parameters, smoking, alcohol and coffee consumption, hypertension, education level, physical activity, and sleep duration from European‐descent genome‐wide association studies. The genetic associations with blepharoptosis were derived from the FinnGen Study, included 1349 cases and 203,231 controls. Inverse‐variance weighted was mainly used for statistical analyses.

Results: Genetically predicted leisure screen time (odds ratio per 1 standard deviation [95% confidence interval], 1.79 [1.23–2.60]) was associated with higher odds of blepharoptosis. There is indicative evidence that a higher body mass index (BMI) and current smoking are associated with higher blepharoptosis odds, whereas physical activity and smoking cessation are associated with lower blepharoptosis odds. No association with other genetically predicted factors was observed.

Conclusion: The Mendelian randomization study identified current smoking status and BMI as causal risk factors, and smoking cessation as protective factor for blepharoptosis. This suggests that these modifiable risk factors are crucial targets for preventing blepharoptosis.

Graphical abstract

Keywords

blepharoptosis / causal association / lifestyle behavior / Mendelian randomization

1 Introduction

Blepharoptosis, also called ptosis, is one of the most prevalent eyelid malposition disorders, characterized by an abnormally low‐lying upper eyelid rest 1–2 mm below the superior corneoscleral limbus in primary gaze. The estimated prevalence of ptosis (congenital and acquired) varies between 0.79 and 1.99 per 10,000 people in different populations [1, 2]. Ptosis can occur through multiple mechanisms: aponeurotic, neurogenic, traumatic, congenital, mechanical, or myogenic [3]. The sagging of the upper eyelid can impact a patient's appearance and compromise visual function to various degrees. Associated complaints include fatigue of the muscles in the forehead, neck soreness, and appearance anxiety, thereby having a negative impact on the quality of life [4]. Ptosis is typically corrected by surgery performed under local anesthesia; however, the repair is challenging with potential complications [5].

Risk factors including aging, obesity, hypertension, diabetes, cataracts, and possibly smoking are reported to increase the prevalence of blepharoptosis according to previous observational studies [6–9]. However, due to unobserved confounding, reverse causality, and other biases that may impede causal inference in these associations within observational studies, the causal nature of the aforementioned risk factors' associations with ptosis remains uncertain.

Mendelian randomization (MR) design employs genetic variants as instrumental variables for exposure to determine whether an observational association between risk factors and a disease is consistent with a causal effect [10]. Further, the MR design can substantially reduce the likelihood of reverse causality, as genetic variants are immutable by the onset or progression of the disease [11]. In this study, we conducted MR analyses to evaluate the relationships between 16 risk factors and lifestyle behaviors with the risk of ptosis, thereby diminishing reverse causality and minimizing confounding factors.

2 Methods

2.1 Two‐Sample MR Approach

We employed a two‐sample MR design in this study [12]. MR is a type of genetic instrumental variable analysis that utilizes single nucleotide polymorphisms (SNPs) as instrumental variables for the risk factors under investigation. SNPs are randomly assigned during meiosis, and, as a result, they are not susceptible to reverse causation bias. This approach is based on three key assumptions regarding SNPs: (1) associated with the exposure of interest; (2) independent of environmental confounders; (3) affect the outcome solely via the exposure [13], which enables the estimation and testing of the causal effect of the exposure on the outcome [14, 15]. All data utilized in this MR study are publicly accessible. Ethical approval and informed consent were secured in each of the original study.

2.2 Data Sources for and Selection of Genetic Instruments

We conducted searches of genome‐wide association studies (GWAS) to identify leading SNPs to serve as genetic instrumental variables. In cases where multiple studies were found for a single trait, only the largest study with replication was selected for use. The UK Biobank is a cohort study that enrolled more than 500,000 men and women from the general population of the UK between 2006 and 2010 [16]. Leisure screen time was defined according to the original GWAS as self‐reported non‐occupational screen‐based sedentary behavior, including television viewing, recreational computer use, and other non‐work‐related electronic device use.

SNPs associated with 16 modifiable factors in European population were selected through the following procedures: (1) Association reaching the genome‐wide significance threshold (p < 5 × 10⁻⁸) were obtained from the relevant GWASs; (2) Linkage disequilibrium among the SNPs was estimated using 1000 Genomes European panel as the reference population; (3) Independent SNPs, defined by r² < 0.001 and a physical distance exceeding 10,000 kb, were used in this study.

2.3 Ptosis Data Sources

We obtained the genetic associations of the instrumental variables with ptosis from the FinnGen Study (eighth release). The FinnGen Study is a Finnish, nationwide meta‐analysis of GWAS encompassing 20 cohorts and biobanks [17], and the cohort baseline data were electronically linked nationwide. The FinnGen Study comprised 3079 individuals with ptosis (1785 male patients and 1294 female patients), and 203,231 individuals without ptosis. Ptosis was defined as the presence of eyelid dropping. Information on ptosis was sourced from discharge registries utilizing the International Classification of Diseases, 10th Revision (ICD‐10) diagnostic code H02.4 for ptosis. The GWAS accounted for variable such as age, sex, 10 principal components, and genotyping batch.

2.4 Statistical Analyses

SNPs were harmonized with the outcome dataset based on chromosome and position.

The inverse variance weighted (IVW) method served as the primary statistical analysis approach, complemented by four sensitivity analyses. These included the weighted median [18], MR‐Egger [19], and MR‐PRESSO [20] methods, which were employed to assess the consistency of associations and to identify and correct for horizontal pleiotropy. The random‐effect IVW was used to incorporate an overdispersion parameter into the variance calculation, thereby accounting for heterogeneity. The weighted median estimator was used to yield consistent and robust estimates, under the assumption that up to 50% of the weight could originate from invalid instrumental variables. MR‐Egger regression was applied to identify potential directional pleiotropy. MR‐PRESSO is capable of producing corrected estimates by detecting horizontal pleiotropy by removing outliers. Cochran's Q statistic and I² were employed to test the heterogeneity among SNP estimates from different genetic variants [21]. The p‐value obtained from the intercept test of MR‐Egger regression was utilized to gauge horizontal pleiotropy. F‐statistic of at least 10 is sufficient to minimize potential weak instruments bias [22], calculated using the formula: F = R²/(1 − R²) × (N − k − 1)/k, where R² represents the proportion of variance explained, N is the total sample size of exposure data and k is the number of IVs. For R² estimation, we used add_rsq (dat) from two‐sample MR. Steiger filtering was used to identify the stronger bidirectional effects, which should account for more variance in the exposure than in the outcome, and remove those reverse causal IVs that bias the MR estimate. The mean effect estimate was calculated separately for each outcome database using the Wald ratios (gene‐outcome [log odds ratio]/gene‐exposure associations) determined for each instrumental variable [23]. These estimates were presented as odds ratios (ORs) for ptosis risk corresponding to a unit change in each instrumental variable. Additionally, we conducted multivariable MR analysis to distinguish the effects of associated variables on ptosis. The results were adjusted for multiple hypothesis testing using the Benjamini and Hochberg false discovery rate (FDR) method, with the significance threshold established at FDR‐corrected p‐values < 0.05 [24]. Associations with p < 0.05 that did not meet the FDR‐controlled threshold were noted as indicative of a potential association.

All the analyses were conducted using R 3.6.1 version.

3 Results

Inclusion of independent SNPs consisted 114 (used SNPs/identified SNPs) for type 2 diabetes [25], 62 SNPs for fasting insulin [26], 69 SNPs for glycated hemoglobin levels (HbA1c measurement) [26], 140 SNPs for body mass index (BMI) [27], 32 SNPs for smoking initiation [28], 56 SNPs for smoking heaviness [29], 36 SNPs for alcohol consumption [28], 37 SNPs for coffee consumption [30], 7 SNPs for vigorous physical activity [31], 4 SNPs for sedentary behavior [32], 94 SNPs for leisure screen time [33], 64 SNPs for sleep duration (ukb‐b‐4424 [16]) (Table 1 and Figure 1).

Higher genetically predicted BMI was linked to an elevated risk of ptosis. The OR for leisure screen time was 1.79 (95% confidence interval [CI] = 1.23–2.60; p = 0.00235) for one standard deviation (SD) increase in leisure screen time. There were suggestive associations for genetically predicted physical exercise (OR per 1 SD increase = 0.0148; 95% CI = 0.8 × 10⁻²–2.74 × 10⁻²; p = 0.0149), BMI (OR = 1.50; 95% CI = 1.07–2.10; p = 0.0173) and smoking cessation (OR for 1 SD decrease = 0.10; 95% CI = 0.08–0.14; p = 0.0368). No association was observed between genetically predicted alcohol consumption, coffee consumption, type 2 diabetes, pre‐diabetic status, hypertension, or sleep duration and the risk of ptosis in the primary analysis (Figure 2). The association remained significant in the multivariable MR analysis after adjusting for genetically predicted leisure screen time (OR = 1.79; 95% CI = 1.23–2.60; p = 2.35 × 10⁻³), but not associated with physical exercise, BMI, or smoking cessation (Figure 3).

The F statistic for the instrumental variables and the estimated power for all analyses are presented in Supporting Information S1: Table S1. All F statistics for the overall instruments exceeded 10, signifying a robust strength of the genetic instruments utilized. The power was adequate for the analysis of BMI, type 2 diabetes, fasting glucose, HbA1c, hypertension, and cognitive education, but low for the other exposures under study (Supporting Information S1: Table S1). The intercept term in MR‐Egger regression servers as an indication for the presence of directional horizontal pleiotropy. No significant intercept was observed in the MR‐Egger regression for exposures to outcome (Supporting Information S1: Table S2). Additionally, no significant heterogeneity was noted in the IVW and MR‐Egger analysis (Supporting Information S1: Table S1). Steiger test showed a correlation between SNP detection and outcome is greater than that of exposure (Supporting Information S1: Table S1).

4 Discussion

Lifestyle behaviors have been found to be genetically predicted predisposition to multiple diseases [34–36]. This MR study revealed that a higher genetically predicted leisure screen time was linked to an increased ptosis risk. There were suggestive associations of a genetically predicted higher BMI with an elevated ptosis risk, as well as physical exercise, and smoking cessation with a lower risk of ptosis. The association for leisure screen time was maintained after accounting for genetically predicted smoking, BMI, and physical exercise. There was inadequate evidence to establish a causal relationship with other modifiable risk factors and ptosis. From a clinical prevention perspective, our findings suggest that maintaining healthy body weight, reducing prolonged recreational screen exposure, reducing smoking and promoting regular physical activity may help reduce blepharoptosis risk. These modifiable behaviors may be incorporated into preventive counseling, particularly among high‐risk populations.

The association we observed between BMI and ptosis is likely causal, given that most sensitivity analyses were consistent and showed no signs of violating the assumption of MR. Moreover, our findings aligned with previous prospective observational studies [6, 7, 9, 37, 38]. Prior prospective observational research [7] has indicated a potential nonlinear association between obesity parameters and ptosis risk, whereas moderate drinkers did not show an increased risk compared with abstainers. Since our MR study was not structured to detect nonlinear associations, we can only infer that high BMI is probably a causal risk factor for ptosis.

The observed associations of genetic predisposition to leisure screen time, physical activity and an increased ptosis risk were not revealed in any prior randomized intervention studies. However, this could be supported by a randomized intervention study that has shown that urban residency was significantly associated with ptosis [37].

The observational literature on the association between smoking and incident ptosis is conflicting. One previous study reported current smoking was possibly involved in the origin of ptosis [9]. However, recently two other studies found that smoking is not associated with an increased risk of developing ptosis [39, 40]. Despite the ambiguity surrounding the evidence linking smoking to ptosis, smoking should still be discouraged, as it remains a potential risk factor for ptosis.

This MR study did not corroborate the association with prediabetic status, hypertension, alcohol assumption, and education level documented by observational studies [6, 7]. This may suggest that the associations identified in observational studies could be reflective of confounding factors or reverse causation bias. On the other hand, it is also possible that our null findings resulted from insufficient statistical power, given the relatively low variance explained for some of these risk factors. Overall, we cannot draw any definitive conclusions about the causal role of these modifiable risk factors in ptosis based on our MR analysis.

After adjustment in multivariable MR analyses, fewer exposures remained statistically significant. This may reflect overlapping biological pathways among metabolic traits, collinearity between correlated exposures, reduced statistical power in multivariable models, and mediation effects whereby certain exposures influence blepharoptosis indirectly through obesity‐related metabolic pathways.

Ptosis is marked by a multifactorial and intricate pathophysiology [41]. Blepharoptosis may sometimes occur independently but can also serve as a clinical indicator of other health issues, such as neurological, immunological, cerebral, and vascular disorders. Lifestyle behaviors might interfere with a variety of pathophysiological processes as it does in hypertension, nonalcoholic fatty liver disease, and gastroesophageal reflux disease [34–36]. For example, obesity is related to skin sagging, loss of dermal elasticity, circulation disorder, and impaired microvascular blood flow to the eyelids [42, 43]. Physical exercise could possibly alleviate oxidative stress in the pathological study of muscle structures within the levator aponeurosis in cases of blepharoptosis [44]. Furthermore, there are several factors that link smoking to ptosis, such as disrupting the maintenance of delicate connective tissue structures, and various pathologies through its impact on the immune‐inflammatory system. The positive effect of less leisure screen time on ptosis risk might be due to less irritation of the eyes. Additionally, other factors such as having more resources to sustain a healthy lifestyle and accessing healthcare could also play a role.

Observational studies include unmeasured confounders or reverse causation leading to possible biases. Therefore, we rely on MR to minimize the possibility of these biases [45]. However, MR studies may yield a confounded estimate from MR and could potentially result in biased causal estimates [46] due to pleiotropy bias (the association of genetic variants with multiple variable) [47]. While incorporating multiple variants in MR analysis generally enhances statistical power, it also increases the risk of including pleiotropic genetic variants that do not qualify as valid IVs [48]. Consequently, we employed three distinct approaches in our study to bolster confidence in these associations. Our findings lend further support to prior observational studies which have identified a link between lifestyle behaviors and ptosis. However, the present results may offer insight into the underlying mechanisms through which lifestyle behaviors impact the risk of ptosis.

Limitations need consideration when interpreting our results. First, for specific exposures such as alcohol consumption and physical exercise, the nonlinear association could not be assessed in the current MR analysis relying on summary‐level genetic statistics. Suppose severe obesity may confer substantially greater risk than moderate overweight. Future studies using individual‐level datasets such as UK Biobank or FinnGen are warranted to investigate dose‐response and non‐linear relationships. Observational studies have documented that the prevalence of various lifestyle and metabolic factors, as well as ptosis, varies by age or sex [7, 35]. We were unable to perform age‐ or sex‐stratified analyses in the present MR study based on summary‐level data, and this requires further investigation on potential effect modification using individual‐level datasets. Second, our analysis included a relatively small number of SNPs as IVs for some exposures, which may have limited our power to detect an association. Another important limitation relates to statistical power for certain behavioral exposures. Although all instrumental variables demonstrated adequate strength (all F statistics > 10), several traits were instrumented by relatively few SNPs and explained only a small proportion of exposure variance. This may have reduced our ability to detect modest causal effects and contributed to false‐negative findings. In addition, sensitivity analyses for pleiotropy may be less stable when only a limited number of variants are available. Therefore, null findings for these exposures should be interpreted cautiously and require validation in larger GWAS datasets. Third, the current study was based on participants of European ancestry. Genetic architecture may vary across populations due to differences in allele frequencies, linkage disequilibrium patterns, gene‐environment interactions, and lifestyle distributions. Caution is needed when generalizing beyond European populations. Future trans‐ancestry Mendelian randomization studies involving East Asian, African, and other ethnically diverse populations are needed to validate the generalizability of these findings. Also, blepharoptosis is a heterogeneous condition that includes congenital, aponeurotic, neurogenic, myogenic, and mechanical subtypes with distinct etiologies. Subtype‐stratified analyses were not feasible in FinnGen dataset. Future studies using clinically phenotyped cohorts are needed. Furthermore, there is currently no clinical intervention to support our findings, and validation of our results through observational studies is needed.

5 Conclusion

Addressing ptosis and lifestyle‐related risk factors identified in this study will aid in the prevention of ptosis, consequently, reduce the associated disease burden and mortality. These findings highlight several modifiable lifestyle and metabolic factors as potential targets for primary prevention strategies for blepharoptosis and may inform future public health interventions and patient education efforts.

References

[1] G. J. Griepentrog, N. N. Diehl, and B. G. Mohney, “Incidence and Demographics of Childhood Ptosis,” Ophthalmology 118, no. 6 (2011): 1180–1183.

[2] A. Y. Nemet, O. Segal, M. Mimouni, and S. Vinker, “Associated Morbidity of Pediatric Ptosis: A Large, Community Based Case‐Control Study,” Graefes Archive for Clinical and Experimental Ophthalmology 252, no. 9 (2014): 1509–1514.

[3] J. M. Lim, J. H. Hou, R. M. Singa, V. K. Aakalu, and P. Setabutr, “Relative Incidence of Blepharoptosis Subtypes in an Oculoplastics Practice at a Tertiary Care Center,” Orbit 32, no. 4 (2013): 231–234.

[4] C. W. McInnes and M. Lee‐Wing, “Eyelid Ptosis,” Canadian Medical Association Journal 187, no. 14 (2015): 1074.

[5] B. R. Frueh, D. C. Musch, and H. M. McDonald, “Efficacy and Efficiency of a Small‐Incision, Minimal Dissection Procedure Versus a Traditional Approach for Correcting Aponeurotic Ptosis,” Ophthalmology 111, no. 12 (2004): 2158–2163.

[6] M. H. Kim, J. Cho, D. Zhao, et al., “Prevalence and Associated Factors of Blepharoptosis in Korean Adult Population: The Korea National Health and Nutrition Examination Survey 2008–2011,” Eye (London) 31, no. 6 (2017): 940–946.

[7] J. S. Paik, S. K. Jung, K. D. Han, S. D. Kim, Y. M. Park, and S. W. Yang, “Obesity as a Potential Risk Factor for Blepharoptosis: The Korea National Health and Nutrition Examination Survey 2008–2010,” PLoS One 10, no. 7 (2015): e0131427.

[8] J. Bacharach, W. W. Lee, A. R. Harrison, and T. F. Freddo, “A Review of Acquired Blepharoptosis: Prevalence, Diagnosis, and Current Treatment Options,” Eye (London) 35, no. 9 (2021): 2468–2481.

[9] L. C. Jacobs, F. Liu, I. Bleyen, et al., “Intrinsic and Extrinsic Risk Factors for Sagging Eyelids,” JAMA Dermatology 150, no. 8 (2014): 836–843.

[10] S. Burgess and S. G. Thompson, Mendelian Randomization: Methods for Using Genetic Variants in Causal Estimation (Chapman and Hall/CRC, 2015).

[11] A. E. Connor, A. V. Khera, and S. Kathiresan, “Mendelian Randomization,” JAMA 318, no. 19 (2017): 1925–1926.

[12] V. Didelez and N. Sheehan, “Mendelian Randomization as an Instrumental Variable Approach to Causal Inference,” Statistical Methods in Medical Research 16, no. 4 (2007): 309–330.

[13] J. B. Frederick and X. Zhou, “Statistical Methods for Mendelian Randomization in Genome‐Wide Association Studies: A Review,” Computational and Structural Biotechnology Journal 20 (2022): 2338–2351.

[14] D. A. Lawlor, R. M. Harbord, J. A. C. Sterne, N. Timpson, and G. Davey Smith, “Mendelian Randomization: Using Genes as Instruments for Making Causal Inferences in Epidemiology,” Statistics in Medicine 27, no. 8 (2008): 1133–1163.

[15] T. M. Palmer, J. R. Thompson, M. D. Tobin, N. A. Sheehan, and P. R. Burton, “Adjusting for Bias and Unmeasured Confounding in Mendelian Randomization Studies With Binary Responses,” International Journal of Epidemiology 37, no. 5 (2008): 1161–1168.

[16] Biobank UK, “UK Biobank: Protocol for a Large‐Scale Prospective Epidemiological Resource,” accessed August, 2020.

[17] FinnGen, “FinnGen: Documentation of R8 Release,” accessed February 20, 2022.

[18] J. Bowden, G. Davey Smith, P. C. Haycock, and S. Burgess, “Consistent Estimation in Mendelian Randomization With Some Invalid Instruments Using a Weighted Median Estimator,” Genetic Epidemiology 40, no. 4 (2016): 304–314.

[19] J. Bowden, G. Davey Smith, and S. Burgess, “Mendelian Randomization With Invalid Instruments: Effect Estimation and Bias Detection Through Egger Regression,” International Journal of Epidemiology 44, no. 2 (2015): 512–525.

[20] M. Verbanck, C. Y. Chen, B. Neale, and R. Do, “Detection of Widespread Horizontal Pleiotropy in Causal Relationships Inferred From Mendelian Randomization Between Complex Traits and Diseases,” Nature Genetics 50, no. 5 (2018): 693–698.

[21] M. F. Greco, C. Minelli, N. A. Sheehan, and J. R. Thompson, “Detecting Pleiotropy in Mendelian Randomisation Studies With Summary Data and a Continuous Outcome,” Statistics in Medicine 34, no. 21 (2015): 2926–2940.

[22] S. Burgess, S. G. Thompson, and CRP CHD Genetics Collaboration, “Avoiding Bias From Weak Instruments in Mendelian Randomization Studies,” International Journal of Epidemiology 40, no. 3 (2011): 755–764.

[23] J. R. Staley and S. Burgess, “Semiparametric Methods for Estimation of a Nonlinear Exposure‐Outcome Relationship Using Instrumental Variables With Application to Mendelian Randomization,” Genetic Epidemiology 41, no. 4 (2017): 341–352.

[24] Y. Benjamini and D. Yekutieli, “The Control of the False Discovery Rate in Multiple Testing Under Dependency,” Annals of Statistics 29, no. 4 (2001): 1165–1188.

[25] A. Xue, Y. Wu, Z. Zhu, et al., “Genome‐Wide Association Analyses Identify 143 Risk Variants and Putative Regulatory Mechanisms for Type 2 Diabetes,” Nature Communications 9, no. 1 (2018): 2941.

[26] J. Chen, C. N. Spracklen, G. Marenne, et al., “The Trans‐ancestral Genomic Architecture of Glycemic Traits,” Nature Genetics 53, no. 6 (2021): 840–860.

[27] T. J. Hoffmann, H. Choquet, J. Yin, et al., “A Large Multiethnic Genome‐Wide Association Study of Adult Body Mass Index Identifies Novel Loci,” Genetics 210, no. 2 (2018): 499–515.

[28] M. Liu, Y. Jiang, R. Wedow, et al., “Association Studies of up to 1.2 Million Individuals Yield New Insights Into the Genetic Etiology of Tobacco and Alcohol Use,” Nature Genetics 51, no. 2 (2019): 237–244.

[29] F. Koskeridis, E. Evangelou, S. Said, et al., “Pleiotropic Genetic Architecture and Novel Loci for C‐Reactive Protein Levels,” Nature Communications 13, no. 1 (2022): 6939.

[30] J. B. Cole, J. C. Florez, and J. N. Hirschhorn, “Comprehensive Genomic Analysis of Dietary Habits in UK Biobank Identifies Hundreds of Genetic Associations,” Nature Communications 11, no. 1 (2020): 1467.

[31] Y. C. Klimentidis, D. A. Raichlen, J. Bea, et al., “Genome‐Wide Association Study of Habitual Physical Activity in Over 377,000 UK Biobank Participants Identifies Multiple Variants Including CADM2 and APOE,” International Journal of Obesity 42, no. 6 (2018): 1161–1176.

[32] A. Doherty, K. Smith‐Byrne, T. Ferreira, et al., “GWAS Identifies 14 Loci for Device‐Measured Physical Activity and Sleep Duration,” Nature Communications 9, no. 1 (2018): 5257.

[33] Z. Wang, A. Emmerich, N. J. Pillon, et al., “Genome‐Wide Association Analyses of Physical Activity and Sedentary Behavior Provide Insights Into Underlying Mechanisms and Roles in Disease Prevention,” Nature Genetics 54, no. 9 (2022): 1332–1344.

[34] S. van Oort, J. W. J. Beulens, A. J. van Ballegooijen, D. E. Grobbee, and S. C. Larsson, “Association of Cardiovascular Risk Factors and Lifestyle Behaviors With Hypertension: A Mendelian Randomization Study,” Hypertension 76, no. 6 (2020): 1971–1979.

[35] S. Yuan, J. Chen, X. Li, et al., “Lifestyle and Metabolic Factors for Nonalcoholic Fatty Liver Disease: Mendelian Randomization Study,” European Journal of Epidemiology 37, no. 7 (2022): 723–733.

[36] S. Yuan and S. C. Larsson, “Adiposity, Diabetes, Lifestyle Factors and Risk of Gastroesophageal Reflux Disease: A Mendelian Randomization Study,” European Journal of Epidemiology 37, no. 7 (2022): 747–754.

[37] B. J. Cho, Y. J. Choi, M. C. Shin, S. W. Yang, and M. J. Lee, “Prevalence and Risk Factors of Childhood Blepharoptosis in Koreans: The Korea National Health and Nutrition Examination Survey,” Eye (London) 34, no. 9 (2020): 1585–1591.

[38] C. Lee, S. M. Jeong, and H. S. Sa, “Correlation Between Body Composition and Blepharoptosis in Adults,” Eye (London) 35, no. 7 (2021): 1999–2007.

[39] D. Bosco, R. Costa, M. Plastino, et al., “Glucose Metabolism in the Idiopathic Blepharoptosis: Utility of the Oral Glucose Tolerance Test (OGTT) and of the Insulin Resistance Index,” Journal of the Neurological Sciences 284, no. 1–2 (2009): 24–28.

[40] N. Satariano, M. S. Brown, S. Zwiebel, and B. Guyuron, “Environmental Factors That Contribute to Upper Eyelid Ptosis: A Study of Identical Twins,” Aesthetic Surgery Journal 35, no. 3 (2015): 235–241.

[41] K. Patel, S. Carballo, and L. Thompson, “Ptosis,” Disease‐a‐Month 63, no. 3 (2017): 74–79.

[42] M. Shirado, “Dyslipidaemia and Age‐Related Involutional Blepharoptosis,” Journal of Plastic, Reconstructive & Aesthetic Surgery 65, no. 6 (2012): e146–e150.

[43] T. Ezure and S. Amano, “Influence of Subcutaneous Adipose Tissue Mass on Dermal Elasticity and Sagging Severity in Lower Cheek,” Skin Research and Technology 16, no. 3 (2010): 332–338.

[44] S. Kase, M. Noda, H. Yoshikawa, T. Yamamoto, K. Ishijima, and S. Ishida, “Oxidative Stress in the Levator Aponeurosis in Asian Involutional Blepharoptosis,” Ophthalmic Plastic and Reconstructive Surgery 30, no. 4 (2014): 290–294.

[45] G. D. Smith and S. Ebrahim, “Mendelian Randomization: Genetic Variants as Instruments for Strengthening Causal Inference in Observational Studies,” in Biosocial Surveys, ed. M. Weinstein, J. W. Vaupel, and K. W. Wachter (National Academies Press, 2008), 336–366.

[46] A. B. Paaby and M. V. Rockman, “The Many Faces of Pleiotropy,” Trends in Genetics 29, no. 2 (2013): 66–73.

[47] J. R. Thompson, C. Minelli, J. Bowden, et al., “Mendelian Randomization Incorporating Uncertainty About Pleiotropy,” Statistics in Medicine 36, no. 29 (2017): 4627–4645.

[48] G. D. Smith and S. Ebrahim, “Mendelian Randomization: Prospects, Potentials, and Limitations,” International Journal of Epidemiology 33, no. 1 (2004): 30–42.

Rights & permissions

2026 The Author(s). Eye & ENT Research published by John Wiley & Sons Australia, Ltd on behalf of Higher Education Press.