Smoking and Risk of Age-Related Cataract: A Meta-Analysis

Abstract

Purpose.: We conducted a meta-analysis to evaluate the relationship between smoking and age-related cataract (ARC).

Methods.: Eligible studies were identified via computer searches and reviewing the reference lists of the key articles. The summary relative risk ratio (RR) or odds ratio (OR) and 95% confidence interval (CI) were calculated. Study-specific risk estimates were pooled using a random-effects model. Meta-regression to assess heterogeneity by several covariates and subgroup analysis on ARC types were performed.

Results.: A total of 13 prospective cohort and eight case-control studies met our inclusion criteria. Ever smoking was statistically significantly associated with increased risk of ARC among cohort studies (OR 1.41, 95% CI 1.23–1.62) and case-control studies (OR 1.57, 95% CI 1.20–2.07). In subgroup analysis, ever smoking exhibited a positive relationship with nuclear cataract (NC; OR 1.66, 95% CI 1.46–1.89) and a marginally significant relationship with posterior subcapsular cataract (OR 1.43, 95% CI 0.99–2.07) in cohort studies. Similar results were found in case-control studies (NC OR 1.86, 95% CI 1.47–2.36; posterior subcapsular cataract OR 1.60, 95% CI 0.97–2.65). Current smokers were at higher risk of ARC than past smokers. No association between smoking and cortical cataract was observed.

Conclusions.: The overall current literature suggests that smoking was associated with increased risk of ARC, especially NC. Further efforts should be made to confirm these findings and clarify the underlying biological mechanisms.

Introduction

Age-related cataract (ARC) remains the leading cause of blindness in the world. ^1,2 According to the World Health Organization's (WHO) latest assessment, ARC is responsible for 51% of world blindness, which represents about 20 million people. ² Although cataracts can be removed surgically and replaced by an artificial intraocular lens to restore sight, many people remain blind from cataracts due to inadequate surgical services and high surgery expenses. ³ Furthermore, with the rapidly aging population, cataract-induced visual dysfunction and blindness are on the increase. ⁴ These diseases are becoming a significant social problem all over the world. ⁵ Therefore, preventing cataracts or delaying the progression to visual disability carries the potential for significant benefits, and the financial as well as clinical burden of the disease could be abridged. Thus, identifying modifiable risk factors for cataract is important and may help to establish preventive measures. A cluster of epidemiologic studies in populations from around the world already has evaluated and identified several factors associated with an increased risk of ARC, ^6–8 such as age, sunlight exposure, and alcohol consumption. 

Smoking is a well-known risk factor for a wide range of diseases, such as vascular disease, lung cancer, and chronic obstructive pulmonary disease. ^9–11 Tobacco smoke contains hundreds of different substances, including nicotine, free radicals, and carbon monoxide, which can increase oxidative stress and have an important role in the pathogenesis of ARC. ^12,13 A number of prior epidemiologic studies suggest that smoking was associated with an increased risk of ARC, ^6,7,14–20 particularly nuclear cataract (NC), but quantitative evidence to certify this association still is lacking. Therefore, our meta-analysis was performed to address this gap in the evidence, providing robust evidence for the association. 

We updated and assessed quantitatively the effects of ever smoking, current smoking, and past smoking on the risk of ARC from the cohort and case-control studies. We also investigated whether the association between smoking and ARC risk differed by types of ARC. 

Methods

Search Strategy

We identified relevant publications in the MEDLINE database using PubMed (http://www.ncbi.nlm.nih.gov/pubmed), Web of Science (http://apps.webofknowledge.com/WOS), and the Cochrane Library (http://www.thecochranelibrary.com/view/0/index.html) up to August 2011. Search terms included “smoking,” “tobacco,” “cigarette,” or “lifestyle” combined with “cataract” or “lens opacities.” The titles and abstracts were scanned to exclude any clearly irrelevant studies. The full texts of the remaining articles were read to determine whether they contained information on the topic of interest. Furthermore, references in the retrieved publications, as well as those in previous reviews, ^21–24 were checked for any other pertinent studies. 

Study Selection (Fig. 1)

For the purpose of meta-analysis, eligible studies had to fulfill all of the following inclusion criteria: (1) case–control or cohort study published as an original article, (2) papers reported in English between 1980 and August 2011, (3) estimation of the relationship between active smoking and the risk of ARC expressed as odds ratio (OR) or relative risk (RR) with their corresponding 95% confidence intervals (CIs), and (4) adjustment made for potential risks, at least age, or sufficient information allowing us to compute them. We also excluded studies that were limited to non-generalizable patients, including two studies of ARC in diabetic patients. ^25,26 In studies with overlapping patients or controls, only the latest or the most complete were included. Any study with inconsistent or erroneous data was excluded. Meeting abstracts with insufficient data or unpublished reports were not considered. 

Figure 1. 

 

Flowchart of study selection process. HR, hazard ratios.

Data Extraction

Information from the included studies was extracted independently by two researchers (JH and CW). Conflicting evaluations were resolved by discussion. If a consensus still could not be reached, the senior investigator (JY) made the final decision. We extracted the name of the first author, year of publication, country in which the study was conducted, sample size, mean age, outcomes definitions and grading, smoking exposure status, mean follow-up time for prospective cohort studies, and covariates included in the final adjusted models. Our primary analysis compared the risk of ARC in ever smokers to never smokers. Several studies did not report an overall RR or OR for ever smoking, but separate adjusted odds ratios of several smoking consumption strata. For these studies, we abstracted the adjusted odds ratio comparing the highest category of smoking consumption with the lowest category (reference group). In addition, summary estimates also were calculated according to smoking status (past and current smoking) and ARC subtypes (NC, cortical cataract, and posterior subcapsular cataract). If a study provided several risk estimates, the most completely adjusted estimate was extracted. 

Statistical Analysis

The OR was used as a measure of the relative risk for all studies, and the RR estimates were log-transformed. The data from individual studies were pooled by use of the random-effect model with the DerSimonian-Laird method, ²⁷ which considers within-study and between-study variation. We performed subgroup analyses based on smoking status (past and current smoking) and ARC subtypes (NC, cortical cataract, and posterior subcapsular cataract). Sensitivity analyses were performed, excluding the two studies where the outcome was assessed by self-report and chart review. ^28,29  

The Q-statistic and I² score were used to assess the between-study heterogeneity of results. ^30,31 Meta-regression analysis was used to assess the heterogeneity in publication year, length of follow-up, and sample size, age, and geographical region. Publication bias assessment was done using the Egger regression asymmetry test ³² and the Begg adjusted rank correlation test. ³³ The statistical software used was Stata/SE 11.0 (Stata Corporation, College Station, TX), and the significance level was set to P < 0.05. 

Results

The search revealed 376 articles, 318 of which were excluded after first-pass review of titles and/or abstract because they were not relevant to the subject of ARC and smoking. Upon closer examination, 37 studies were excluded for the following reasons: 24 articles were cross-sectional studies, ^{6,7,14–20,34–48} five studies did not provide sufficient information to estimate a summary OR and its 95% CIs ^49–52 or a summary OR adjusted at least for age, ⁵³ three studies provided hazard ratios instead of OR or RR, ^54–56 two studies estimated the incidence of ARC among diabetic patients, ^25,26 two studies were updated by Klein et al. ⁵⁷ and Hankinson et al., ⁵⁸ and one study identified cataract subtypes as waterclefts and retrodots. ⁵⁹  

Finally, 13 cohorts and eight case-control studies were evaluated further in this analysis. Two of the cohort articles were analyses at different time points from the same study, ^28,60 but the 13.6-year follow-up study ²⁸ did not reported the relationship between smoking and subtype of ARC; therefore, both articles were included as one study. 

Tables 1 and 2 provide summaries of the study designs and participant characteristics. Not all studies reported every subtype of ARC. Age-related cataract assessments and definitions varied among the studies. Standardized criteria for diagnosis of cataract were used in some studies, while in others cases were diagnosed medically by ophthalmologist or medical record review to identify the case. Likewise, the outcome measure of cataract was not consistent. Many studies used the prevalence or incidence of cataract, but some studies used cataract extraction as the measure of outcome, and found an association between smoking and cataract extraction. 

Table 1. 

 

Prospective Cohort Studies Evaluating the Association between Smoking and Cataract

Source (Publication Yr., Country)	Mean Follow-up Time (yrs.)	Study Period	Population (Sample Size)	Mean Age (yrs.)	ARC Outcome	ARC Definition and Grading	Smoking Exposure Status	Adjusted Variables
Christen et al.60 (1992 USA)	5	1982–1987	Volunteer health professionals (N = 17,824 male only)	40−84	Nuclear PSC Any type	Self-report confirmed by medical record review	Past Current  (≥20 cigarettes/ day)	Age, aspirin, β-carotene treatment assignment, history of diabetes, hypertension, high cholesterol levels, obesity, alcohol use, physical activity, parental history of myocardial infarction.
Christen et al.28 (2000 USA)	13.6	1982–1997	Volunteer health professionals (N = 20,907 male only)	40−84	Any type Extraction	Self-report confirmed by medical record review	Past Current	Age, diabetes, hypertension, BMI, alcohol use, aspirin and β-carotene treatment assignment, physical activity, parental history of myocardial infarction, current multivitamin use, and number of cigarettes smoked
Klein et al.61 (2003 USA)	5	1987–1995	Population-based (N = 3664)	43−84	Nuclear Cortical PSC Extraction	The Wisconsin Cataract Grading system	Current (≥35 pack-years)	Age, sex
Leske et al.64 (1998 USA)	4.6	1989–1993	Clinic-based (N = 764)	≥40	Nuclear Opalescence  increased	LOCS III	Current others	Age, education
Weintraub et al.29 (2002 USA)	16 (female)	1980–1996	Volunteer health professionals (N = 124,690)	>45	Extraction	Self-report confirmed by ophthalmologists or nurses	Current Past	Age, diabetes mellitus, BMI, dietary intake of lutein/zeaxanthin, state of residence at baseline, and 2-year time period
	10 (male)	1986–1996
Mares et al.69 (2010 USA)	6	1994–2004	Population-based (N = 1808 female only)	50−79	Nuclear or extraction	AREDS-SCC	Ever (pack- years ≥7)	Age, iris pigmentation, HEI-1995 score, BMI, pulse pressure, dietary variables, energy
Mukesh et al.67 (2006 Australia)	5	1992–1999	Population-based (N = 3721)	≥40	Cortical Nuclear PSC	The Wilmer cataract grading system	Past Current	Age, sex, diagnosed with arthritis, occupation, country of birth, diabetes
Lindblad et al.66 (2005 Sweden)	5	1997–2002	Population-based (N = 34,595 female only)	49−83	Extraction	The Swedish National Cataract Register	Past Current  >(smokers 15 cigarettes/day)	Age, diabetes, hypertension, steroid medication use, alcohol consumption, vitamin supplement use, BMI, education.
Tan et al.68 (2008 Australia)	10.5	1992–2004	Population-based (N = 2406)	≥49	Nuclear Cortical PSC Extraction	The Wisconsin Cataract Grading System	Ever Past Current (pack-  years ≥36)	Age, sex, hypertension, myopia, diabetes, skin damage, ever use of oral or inhaled steroids.
Delcourt et al.65 (2003 France)	3	1998–2000	Population-based (N = 1736)	≥60	Nuclear Cortical PSC Extraction	LOCS III	Current Past Ever (pack-years  ≥40)	Age, gender and C0, NO0, P0 for cataract surgery, and NO0, P0 for cortical cataract, and C0, P0 for nuclear cataract, and C0, NO0 for posterior subcapsular cataract
Hiller et al. 63 (1997 USA)	12.5	1973–1989	Surviving members of the Framingham Heart Study Cohort (N = 660)	52−80	Nuclear opacity Nonnuclear  opacity	The Taylor and West grading system	Past Current (≥20  cigarettes /day)	Age, sex, education, diabetes
West et al.62 (1995 USA)	5	1985–1990	Volunteer health professionals (N = 442 male only)	≥30	Nuclear	The Wilmer cataract photograph grading system.	Current Past	Age, baseline opacity status, and alcohol use
Zhang et al.70 (2011 China)	5	2001–2006	Population-based (N = 3251)	≥45	Nuclear Cortical PSC Any type Extraction	Modified AREDS-SCC	Ever	Age, gender, education, income, BMI, refractive error, alcohol consumption

 LOCS III, Lens Opacities Classification System III; AREDS-SCC, The Age-Related Eye Disease Study System for Classifying Cataracts; BMI, body mass index.

Table 2. 

 

Case-Control Studies Evaluating the Association between Smoking and Cataract

Source (Published Year, Country)	Designed	Case/Control (Mean Age, yrs.)	Cataract Type	Case Definition	Smoking Exposure Status	Adjusted Variables
Harding et al.71 (1988 UK)	Hospital-based Case-control	124/266 (50−79)	Extraction	Admitted to the Oxford Eye Hospital for cataract extraction	Heavy smokers (with cigarette-year score >1500)	Age, sex
AREDS Research group77 (2001 USA)	Population-based Case-control	NS/4757 (60−80)	Moderate Nuclear Opacity	AREDS-SCC	Current	Age, sex
Theodoropoulou et al.78 (2011 Greece)	Population-based Case-control	314/314 (45−85)	Cortical Nuclear PSC All types	Medically diagnosed and classified in the Ophthalmology Department of the “Attikon” Hospital in Athens, Greece	Current Ex-smoker	Age, sex, education, BMI
Leske et al.72 (1991 USA)	Hospital-based Case-control	945/435 (40−79)	Cortical Nuclear PSC Mixed	LOCS III	Current smoking	Age, sex
Phillips et al.74 (1996 UK)	Hospital-based Case-control	990/858 (NS)	Any type	On the waiting lists of surgeons at Princess Alexandra Eye Pavilion	Smoking	Age, sex
Katoh et al.73 (1993 Japan)	Hospital-based Case-control	212/212 (≥40)	Senile cataract	The criteria of the Japanese Co-operative Cataract Epidemiology Group	Current smoker Ex-smoker	Age, sex
Ughade et al.75 (1998 India)	Hospital-based Group-matched Case-control	262/262 (51−70)	Any type	Sufficiently advanced lens opacity that impaired vision	History of heavy smoking (smoked ≥10 cigarettes/bidis daily for ≥2 yrs.)	Age, low socioeconomic status, sex, illiteracy, history of diabetes, history of diarrhoea, glaucoma, myopia, hypertension, cheap cooking fuel
Ojofeitimi et al.76 (1999 Nigeria)	Hospital-based Case-control	31/31 (60−69)	Any type cataract	Medically diagnosed in the Ophthalmology clinic at Obafemi Awolowo University Teaching Hospitals Complex, Ile-Ife	Past history of smoking	Age, sex

 NS, not stated.

Prospective Cohort Study

Table 1 shows that 12 cohort studies ^{28,29,61–70} were included in the analysis of the association between ever smoking and ARC risk. Of the studies seven were conducted from North America, ^{28,29,61–64,69} two from Europe, ^65,66 two from Australia, ^67,68 and one from Asia. ⁷⁰ There were 11 population-based and one clinic-based study. The mean time to follow-up ranged from 3–16 years. 

We found that ever smoking (OR 1.41, 95% CI 1.23–1.62) was statistically significantly associated with increased risk of ARC. There was significant heterogeneity (I² = 67.8, P = 0.000) among ever smokers (Fig. 2). Further scrutiny found that the heterogeneity among ever smokers shifted from P = 0.000 to P = 0.074 by Q test when two studies with patients with self-reported ARC were excluded (OR 1.32, 95% CI 1.15–1.51, see Supplementary Fig. S1). We used meta-regression analysis to explore the influence of publication year, sample size, and study conducted area. However, none was identified as a possible source of heterogeneity among all the included studies (data not shown). Egger's test suggested no statistically significant asymmetry of the funnel plot (P = 0.437), indicating no evidence of substantial publication bias. 

Figure 2. 

 

Risk estimates of ARC associated with current, past, and ever smoking in prospective cohort studies.

Eight studies were included in the analysis of the association between past smoking and ARC risk, ^{29,60,62,63,65–68} and 10 between current smoking and ARC risk. ^{28,29,61–68} The association was stronger among current smokers (OR 1.47, 95% CI 1.36–1.59) than past smokers (OR 1.19, 95% CI 1.01–1.41, Fig. 2). Statistically significant heterogeneity existed among the eight studies evaluating the past smoking risk of ARC (I² = 63.8, P = 0.007), but not among the 10 studies assessing the current smoking risk of ARC (I² = 4.2, P = 0.402). Removing the studies with patients with self-reported ARC, the estimate was attenuated among past smokers (OR 1.19, 95% CI 0.92–1.53, I² = 58.5, P = 0.034), but essentially unchanged among current smokers (OR 1.34, 95% CI, 1.18–1.53, I² = 0.0, P = 0.492, see Supplementary Fig. S1). 

Figure 3 shows that for the 11 studies ^{29,60–65,67–70} that reported results for ever smoking, there was a strong positive relationship with risk of NC (OR 1.66, 95% CI 1.46–1.89). We found that the association was statistically significant among current smokers (OR 1.78, 95% CI 1.58–2.01), but borderline significant among past smokers (OR 1.19, 95% CI 0.95–1.49). Heterogeneity was detected (P = 0.007) among the seven studies evaluating the past smoking risk of NC. On the contrary, there was no evidence of heterogeneity among all 11 studies (I² = 22.4, P = 0.230) and the nine studies included in the subgroup of current smoking (I² = 5.5, P = 0.389). When the two self-reported ARC studies were excluded (see Supplementary Fig. S2), the estimates among ever smokers and current smokers essentially were unchanged (OR 1.62, 95% CI 1.37–1.91, P = 0.207, and OR 1.67, 95% CI 1.39–2.00, P = 0.387, respectively), while the estimate among past smokers was attenuated (OR 1.19, 95% CI 0.82–1.74, P = 0.027). No publication bias was detected among this subgroup. 

Figure 3. 

 

A forest plot shows risk estimates from cohort studies, estimating the association between NC and different smoking status.

For the five studies that evaluated the association between ever smoking and cortical cataract, there virtually was no association (OR 0.91, 95% CI 0.75–1.11, P = 0.223). ^{61,65,67,68,70} We performed subgroup analysis based on different smoking status. Three studies estimated the association among past smokers (OR 0.93, 95% CI 0.76–1.14, P = 0.934), ^65,67,68 and four studies among current smokers (OR 1.01, 95% CI 0.81–1.26, P = 0.721). ^61,65,67,68 There was no significant association between smoking and cortical cataract risk. The findings were homogeneous across the studies. No publication bias was detected among this subgroup (data not shown). 

Figure 4 shows that seven studies were included in the analysis of the association between ever smoking and posterior subcapsular cataract (PSC) risk, ^{29,60,61,65,67,68,70} five estimated the association among past smokers, ^{29,60,65,67,68} and six among current smokers. ^{29,60,61,65,67,68} The pooled analysis demonstrated a marginally significant association between ever smoking and PSC (OR 1.43, 95% CI 0.99–2.07), while statistically significant association was found among past smokers (OR 1.31, 95% CI 1.05–1.63) and current smokers (OR 1.45, 95% CI 1.05–2.00). The result was heterogeneous and there was no evidence of significant publication bias either with the Egger's test in this subgroup (data not shown). 

Figure 4. 

 

A forest plot shows risk estimates from cohort studies, estimating the association between posterior subcapsular cataract and different smoking status.

Case-Control Studies

Table 2 shows that eight case-control studies were included in the analysis of the association between smoking and ARC risk. ^71–78 Two studies were conducted from North America, ^72,77 three from Europe, ^71,74,78 one from Asia, ⁷³ and two from Africa. ^75,76 Six of the studies were hospital-based, while the other two were population-based. 

We found that ever smoking was associated strongly with increased risk of ARC (OR 1.57, 95% CI 1.20–2.07). As well as the two different statuses of smoking (past smoking OR 1.60, 95% CI 1.02–2.49; current smoking OR 1.55, 95% CI 1.11−2.15, see Supplementary Fig. S3). There was statistically significant heterogeneity among ever smokers (I² = 66.1, P = 0.004) and current smokers (I² = 72.6, P = 0.003). We used meta-regression analysis to explore the influence of publication year, sample size, and study conducted area. However, none was identified as a possible source of heterogeneity among all the included studies. No publication bias was found among all studies (P = 0.322 by Egger's test). 

Because only one case-control study evaluated separate relation between past smoking and each cataract subtype, we pooled estimates for only current smoking in the subgroup analysis. Three studies were included in the analysis of the association between current smoking and NC risk. ^72,77,78 The summary risk was 1.89 (95% CI 1.45–2.45) without being statistically significantly heterogeneous (P = 0.861). Two studies were included in the analysis of the association between current smoking and PSC risk. ^72,78 These same studies also were included in the analysis between current smoking and cortical cataract risk. ^72,78 In this pooled subgroup analysis (see Supplementary Fig. S4), smoking was not associated significantly with the likelihood of cortical cataract (OR 1.09, 95% CI 0.73–1.63, P = 0.939), while the summary risk estimate of PSC was strong (OR 1.89, 95% CI 1.17–3.03, P = 0.512). The findings were homogeneous across the studies. 

Discussion

Our meta-analysis showed that smoking was associated with an increased risk of ARC in cohort and case-control studies. The positive association also was found in the stratified analyses by NC and PSC. The association was stronger among current smokers than past smokers. However, no association between any smoking status and cortical cataract was observed in the cohort and case-control studies. 

The mechanistic actions of smoking on ARC are not fully understood, but several possible biologic mechanisms have been suggested for the association of smoking with ARC. Firstly, oxidative damage appears to have a major role in cataract formation. ^12,23 Smoking causes an additional oxidative challenge through invoking free radical activity, and promoting oxidation and lipid peroxidation. ⁷⁹ On the other hand, smoking may impose indirectly the oxidative stress on the lens through depletion of endogenous antioxidant pools, ^80–82 such as vitamin C, vitamin E, and β-carotene. Secondly, by-products of tobacco contain heavy metals, such as cadmium, lead, and copper, which accumulate in the lens and cause direct toxicity. ^83,84 Thirdly, cyanide and aldehyde levels rise in the blood of smokers, and then aldehydes and isocyanate, which are formed from cyanide, can modify lens proteins, causing lens opacification in vitro. These changes are similar to those seen in human cataracts. ^85–87  

The weaker association with cortical cataract may be a result of different pathophysiologic processes in these situations, suggesting that risk factors may differ for the different cataract types. ⁸⁸ Meanwhile, because the human lens grows throughout life, the lens core is exposed for a longer period to such influences caused by smoking. Another explanation for this relative lack of protection of the lens nucleus from oxidative attack has been provided by Sweeney and Truscott. ⁸⁹ Their studies demonstrated the appearance of a barrier to the inward diffusion of glutathione (GSH), which is the most important antioxidant molecule of the lens. ⁹⁰ The relatively low ratio of GSH to protein-SH in the adult nucleus of the lens, combined with low activity of the GSH redox cycle, makes the nucleus especially vulnerable to oxidative stress. On the contrary, the superficial cortex is supplied better with scavenger molecules to combat oxidative effects. More recently, Mathias et al. ⁹¹ and Donaldson et al. ⁹² hypothesized that micronutrients sift to the center of the lens via a process called electro-osmosis, which may be the cellular and molecular mechanism responsible for lens transparency. Cumulative oxidative damage to any dissipation of the Na⁺ gradient in the core of the lens will reduce nutrients and antioxidants (e.g., GSH) uptake in this region, leading to cataract formation. This may explain why cigarette smoke affects nuclear and posterior cortical opacity more than cortical opacity. 

Our meta-analysis results indicated that current smokers are at higher risk of incident ARC than past smokers. Some previous epidemiological investigations found the same tendency. ^60,63,67,78 One possible reason is that current smokers may have a longer exposure time and higher total cumulative dose of smoking than past smokers. The Swedish Mammography Cohort indicated those who ceased smoking more than 20 years previously had no excess risk of ARC. ⁶⁶ Weintraub et al. pooled results from the Nurses' Health Study and the Health Professionals' Follow-up Study, and indicated that the risk among past smokers decreased with number of years since quitting, but not to the level of never smokers, even 25 or more years after cessation. ²⁹ These two studies suggest that some smoking-related damage in the lens may be reversible on smoking cessation, but the effect of cessation takes some time and may be only partial. 

There was a significant amount of heterogeneity among the studies, likely reflecting differences among study populations, model selection, analytic methodology, exposure assessment, and operational definitions of ARC and its subtypes. We conducted a meta-regression analysis to assess the effect of publication year, study conducted area, study design, primary outcome, and sample size on the heterogeneity. However, none of the confounding factors could explain heterogeneity between the individual studies. We also performed sensitivity analyses excluding the two studies where the outcome was assessed by self-report and chart review. The results showed that the pooled estimates were robust among ever smokers and current smokers. Nonetheless, we failed to find the major source of study heterogeneity in this sensitivity analysis. The presence of heterogeneity indicates the need for consensus definitions for ARC and its subtypes in future studies. 

To interpret our study results properly, it is necessary to understand several limitations. First, only English-language articles that had been published were included. We did not attempt to uncover unpublished observations and did not include studies with insufficient information to estimate an adjusted OR, which could bring publication bias, even though no significant evidence of publication bias was observed in Egger's and Begg's test. Second, we also excluded the cross-sectional studies that did not provide information regarding the temporal relation between ARC and smoking. Third, the assessment of ARC or its subtype varied between studies (Tables 1, 2). Some studies used cataract extraction as the measure of outcome, and found an association between smoking and cataract extraction. However, cataract extraction rates depend on health care provision and access, which if different between smokers and nonsmokers would create bias. In addition, all studies used qualitative and/or quantitative criteria for ARC diagnosis. Qualitative measures, such as levels of lens opacity, introduce the potential of interobserver variation. Fourth, smoking status misclassification is another potential source of bias. The smoking data were self-reported in all included studies. Patients may underestimate or under report their smoking habits, resulting in misclassification of exposure status and inducing bias in estimates of association. Moreover, many prospective cohort studies assess smoking status only at baseline. Smoking status may change during follow-up. Of the 13 cohort studies, seven reassessed smoking status. ^{28,29,58,60–62,69} Finally, the smoking consumption levels in the lowest and highest categories, and the range of smoking consumption varied across studies. These differences may have contributed to the heterogeneity among studies in the analysis of the highest versus lowest categories. 

Conclusions

Our meta-analysis of cohort and case-control studies summarized the risk estimates of smoking and ARC, and provided robust evidence for the association. It helped resolve some of the inconsistencies with smoking and ARC risk, but some do remain. Future research is needed to confirm these findings and resolve the remaining problems. These findings also provide an opportunity for the public health and eye health communities to work actively to educate the public about the impacts of smoking on eye health. Such education will improve quit rates and help to discourage people from starting to smoke. 

Supplementary Materials

pdf S1 

pdf S2 

pdf S3 

pdf S4 

Footnotes

 Supported by project foundation: Zhejiang province Key Lab Fund of China 2011E10006; The Natural Science Foundation of China 81070756; and The Natural Science Foundation of Zhejiang Province of China, Y208396.

Footnotes

 Disclosure: J. Ye, None; J. He, None; C. Wang, None; H. Wu, None; X. Shi, None; H. Zhang, None; J. Xie, None; S.Y. Lee, None   The authors have no financial interest in any product or concept discussed in the article.