Overweight, Obesity, and Risk of Age-Related Macular Degeneration

Abstract

Purpose: The aim of this study was to quantify the relationship between categories of body mass index (BMI) and age-related macular degeneration (AMD) risk in different stages.

Methods: MEDLINE, EMBASE, and ISI Web of Science were searched for all eligible studies on the relationship between BMI and incident early or late AMD. The analyses were based on data extracted from study reports. The pooled relative risks (RRs) with 95% confidence intervals (CIs) were calculated to evaluate the strength of this association, and dose–response relationship was assessed by restricted cubic spline.

Results: Seven prospective cohort studies with 1613 cases identified among 31,151 subjects were included. For overweight, the relationship remained insignificant for its association with both early AMD (RR = 0.92, 95% CI: 0.68–1.15; P = 0.54) and late AMD (RR = 1.09, 95% CI: 0.93–1.25; P = 0.18). A marked 32% increase in the risk of developing late AMD was noted among obese individuals (RR = 1.32, 95% CI: 1.11–1.53, P < 0.01), while obesity showed no significant association with early AMD (RR = 0.91, 95% CI: 0.74–1.08; P = 0.67). Furthermore, elevated BMI showed a linear dose–response relation with AMD risk (P_nonlinearity = 0.17), and the AMD risk increased by 2% (RR = 1.02, 95% CI: 1.01–1.04) for each 1 kg/m² increase in BMI within the overweight and obese BMI ranges.

Conclusions: Excess body weight was weakly associated with increase in the risk of AMD in a dose-dependent fashion, especially for its late stage, indicating that keeping normal body weight and avoiding further weight gain may confer potential protection against this disease.

Age-related macular degeneration (AMD) is a progressive neurodegenerative eye disease that has become the leading cause of irreversible visual impairments in developed countries.¹ Early AMD is characterized clinically by deposits of drusen between the retinal pigment epithelium (RPE) and Bruch's membrane.² This stage can progress to two types of advanced AMD: the dry form includes irregular areas of depigmentation in the retina and geographic atrophy (GA) of the RPE and photoreceptor cells, and the wet form is recognized by abnormal choroidal neovascularization (CNV), exudates, and bleeding in the retina, which leads to severe central vision loss.² It is estimated that this disease is responsible for more than 3 million cases of blindness in the world.³ The projected number of people with the disease is approximately 196 million in 2020 and further to 288 million in 2040, which will bring a heavy socioeconomic burden.⁴ Although anti-vascular endothelial growth factor (anti-VEGF) had a limited effect on the reduction of visual impairment for several wet AMD patients, no available treatment exists for the majority of cases of AMD at present.⁵ Thus, identifying potentially modifiable factors is considered to be far preferable and of major importance to prevent the onset and progression of this disease. 

The pathogenesis of AMD is complicated by both genetic and environmental factors.⁶ Due to the similarities between drusen deposition and the development of atherosclerosis, AMD is supposed to share several of the same risk factors with this vascular disease.⁷ Emerging studies have recognized overweight and obesity as significant risk factors for cardiovascular disease.⁸ Increasing body weight can cause several physical changes, including a higher level of oxidative stress, a higher risk of all inflammatory processes, and imbalance of blood lipids, which is involved in AMD pathogenic mechanisms.^9,10 Previous studies demonstrated that excess body fat may affect the transport and deposition processes of carotenoids from blood to macula, which ultimately led to a decline in the level of macular pigment at the fovea.^11,12 The possible association between excess body weight and AMD has been investigated in several studies, most of which define overweight and obesity by the body mass index (BMI).^13,14 Some studies have found an increased risk with greater BMI, while others have failed to observe this correlation.^13–15 Although dose–response effect is a valuable guide to clinical recommendation, none of the studies has examined the shape of the dose–response relationship of elevated BMI in AMD incidence. Moreover, different mechanisms are suggested to be involved in the pathogenesis at the early and late stages of AMD, and excess weight may have a diverse influence on these two forms.^6,16 

Therefore, we conducted a dose–response meta-analysis of prospective studies to quantitatively evaluate the effect of overweight and obesity on AMD risk. Furthermore, subgroup analyses were also performed to compare roles of excess body weight in different stages of this disease. 

Methods

Search Strategy

We conducted a comprehensive literature search of MEDLINE, EMBASE, and ISI Web of Science for studies up to March 2015. The search terms utilized in this process were (“age-related maculopathy” OR “neovascular AMD” OR “exudative AMD” OR “choroidal neovascularization” OR “geographic atrophy” OR “macular degeneration”) AND (“BMI” OR “body mass index” OR “body size” OR “anthropometry” OR “body weight” OR “overweight” OR “obesity”). No restriction on language of publication was considered. We also manually scanned reference lists of selected articles for pertinent studies and contacted authors and experts in the field for the existence of relevant unpublished studies. 

Study Selection

Three reviewers (Q-YZ, S-SW, L-JT) independently performed an initial screen of titles or abstracts to identify potentially eligible studies, and conducted a full-text review to determine eligibility. Disagreements were resolved by group consensus. To maximize the quality and comparability of the studies, we formulated general inclusion criteria a priori. Research studies obtained were considered eligible if they met the following criteria: Prospective studies evaluated the relationship between BMI and AMD risk; the exposure of interest was BMI with two or more quantitative categories; the outcome was AMD incidence; and the risk estimates including relative risks (RRs) or hazard ratios (HRs) with 95% confidence intervals (CIs) for each BMI category were reported or could be calculated. When more than one publication was identified from the same study, we used the most updated or complete report of that study. 

Data Extraction and Quality Assessment

Three investigators (Q-YZ, S-SW, L-JT) reviewed and extracted the data into a standardized data extraction from each publication. When these individuals failed to concur, a senior investigator (LM) would make the definitive decision for data extraction and recheck it. For each study, we extracted the following information: first author's last name, year of publication, study design, study location, sample size, participants' characteristics (age, sex), duration of follow-up, methods of body size assessment, the categories of BMI, diagnosis method, classification criteria, type of AMD, risk estimates with 95% CI for each category, and covariates adjusted in the statistical analysis. The BMI categories that were closest to the World Health Organization definition of weight status (underweight, ≤18.5 kg/m²; normal weight, 18.5–25 kg/m²; overweight, 25–29.9 kg/m²; and obesity, ≥30 kg/m²) were applied if nonstandard categories were used in individual studies included in this meta-analysis.¹⁷ When different stages and types of AMD were present in a single study, we combined the data into a single group according to the Cochrane recommendation.¹⁸ If the results were reported for two or more multivariable models, we extracted the RRs with 95% CI that reflected the greatest degree of adjustment for potentially confounding variables. Incomplete data presentation was resolved by contact with corresponding authors. 

The quality of the studies was assessed by means of the Newcastle-Ottawa Scale (NOS), which included eight items that were categorized into three aspects: quality of selection, comparability, and quality of exposure.¹⁹ The scoring system provides a summary numeric score of quality ranging from 0 stars to 9 stars. Studies with more than or equal to 6 stars were considered high-quality studies, and those with fewer than 6 stars low-quality studies. Quality assessment was performed independently by three reviewers (Q-YZ, S-SW, L-JT), and any discrepancies were resolved by a senior investigator (LM) through discussion with the three reviewers. 

Statistical Analysis

Pooled RRs with corresponding 95% CIs were calculated to evaluate the strength of relationship between body size and AMD incidence. Heterogeneity of effect sizes in the overall aggregations was assessed using Cochran Q test and I² statistic, which denotes the proportion of between-study heterogeneity. The RRs were pooled using the random-effects model when the P value for heterogeneity was lower than 0.1 or I² values were greater than 50%; otherwise, fixed-effects model was used to compute summary effect. Stratified analyses were performed to investigate the effect of AMD subtypes, BMI and AMD ascertainments, duration of follow-up, source country, AMD classification criteria, and smoking status on the relationship between body size and AMD. Dose–response meta-analysis was performed by the method described by Greenland and Longnecker²⁰ and Orsini et al.²¹ to evaluate a potential trend between BMI levels and AMD risk. The estimated midpoint of each BMI category was utilized if mean or median was not reported in the study. The open-ended categories were assumed to have the same width as their adjacent categories. Because the references of exposure were different across studies, we used centered dose levels (each nonreference dose minus the reference dose within a study) for summarizing dose–response relation.²² In this two-stage analysis, we first estimated restricted cubic spline model with three knots in settled percentiles (25%, 50%, and 75%) of the distribution, assuming the fixed-effects model. Then, the GLST command with the generalized least-squares regression, which required the mid values of BMI in each category, the number of patients and participants, and logarithms of RRs with 95% CIs, was used to carry out the dose–response meta-analysis. P value for nonlinearity was calculated by testing the null hypothesis that the regression coefficients of the second-order spline transformations were equal to zero. Sensitivity analyses were performed by removing one study at a time and recalculating the pooled RR estimates for the remaining studies to evaluate whether the results were robust. Also, to ensure the robustness of the dose–response meta-analysis to different value assignment methods for the open-ended category in each study, we performed sensitivity analysis by assuming the width of open-ended categories to be 1.2, 1.5, and 2.0 times that of adjacent categories. This range of values is broad enough that the robustness of the results of this sensitivity analysis provides convincing evidence for the dose–response relationship. Publication bias was tested by the combination of a funnel plot–based method and use of Begg's test and Egger's test to estimate the number of missing studies and to calculate a corrected RR as if these studies had been present.^23,24 All analyses were performed with STATA version 12.0 (Stata Corp., College Station, TX, USA). A 2-tailed P value of less than 0.05 was considered statistically significant. 

Results

Search Results

Our initial search yielded 2463 potentially relevant abstracts (337 from MEDLINE, 319 from EMBASE, 1804 from ISI Web of Science, and 3 from manual screen of reference lists). After removing duplicates and screening titles and abstracts, 105 studies were retrieved for full text review. Of these, 98 articles were excluded for reasons shown in Figure 1. The remaining 7 articles met the inclusion criteria and were included in the meta-analysis.^{13,14,16,25–28} 

Figure 1

 

Flow diagram of the study selection. BMI, body mass index; AMD, age-related macular degeneration.

Study Characteristics

The main characteristics of the included studies are summarized in Table 1. The study samples varied between 261 and 21,121, and the total number of participants was 31,151 with 1613 reported AMD outcomes. The United States contributed the majority of included studies (5 studies), and the remaining studies were conducted in Denmark (1 study) and Australia (1 study). Five studies comprised both sexes, one consisted entirely of men, and one included only women. There were five population-based studies whereas one study consisted of volunteers. Duration of follow-up ranged from a length of 4 to 14.6 years. Age-related macular degeneration was ascertained based on fundus photographs in six studies, and one study relied on review of medical records. Anthropometric factors were all measured by a trained examiner except for one study using self-reported body size data. All of the studies adjusted for age, and other adjustment factors included sex (six studies), smoking status (five studies), intake of carotenoid and vitamin such as vitamin C and vitamin E (three studies), family history of AMD (two studies), education (two studies), and so on. All studies were of high methodologic quality, with a score range from 6 to 9 stars. 

Table 1

 

Characteristics of Studies Included in This Meta-Analysis

 

Overweight and AMD Subtypes

Two studies reported data only on obesity with AMD; five included studies^{13,14,25,27,28} investigated the relationship between overweight and AMD incidence. Although a slight increase in incidence of AMD was observed among the overweight in most studies, the RR was statistically significant in only one study.²⁸ Compared with normal-weight subjects, overweight was associated with a nonsignificant increase in the risk of AMD (RR = 1.03, 95% CI 0.90–1.17, P = 0.51; I² = 23.0%, P_{heterogeneity} = 0.27; Fig. 2). Two of the five included studies reported on early AMD and three studies reported on late AMD. The relationship remained insignificant both for the association of overweight with early AMD (RR = 0.92, 95% CI: 0.68–1.15, P = 0.54; I² = 0.0%, P_{heterogeneity} = 0.49) and for the association of overweight with late AMD (RR = 1.09, 95% CI: 0.93–1.25, P = 0.18; I² = 0.0%, P_{heterogeneity} = 0.27). 

Figure 2

 

Forest plot on the associations between overweight and age-related macular degeneration. The boxes and lines indicate the relative risks (RRs) and their confidence intervals (CIs) on a log scale for each study. The pooled relative risk is represented by a diamond. The size of the black squares indicates the relative weight of each estimate.

In addition, study and participant characteristics also did not significantly change the shape of association between overweight and AMD risk (Table 2). Sensitivity analyses indicated that the overall finding was not excessively altered by any individual studies. Visual inspection of the funnel plot for the studies evaluating association between overweight and AMD showed no indication of asymmetry. Egger's and Begg's tests revealed the absence of publication bias (Egger's test: P = 0.64, Begg's test: P = 0.81). 

Table 2

 

Stratified Analysis of the Association Between Excess Body Weight and Age-Related Macular Degeneration

 

Obesity and AMD Subtypes

All seven included studies^{13,14,16,25–28} were on the analysis of obesity and risk of AMD. Among these studies, five^{13,16,26–28} found an association between obesity and an increased risk of AMD, and three studies reached statistical significance.^27,28 There was significant heterogeneity (I² = 58.6%, P_{heterogeneity} = 0.03), and results of the random-effects model showed that obese individuals had a slight increased risk of AMD, by 7% (RR = 1.07, 95% CI: 0.94–1.21, P = 0.81; Fig. 3). A statistically significant 32% increase in the risk of developing late AMD was noted among obese individuals (RR = 1.32, 95% CI: 1.11–1.53, P < 0.01; I² = 0%, P_{heterogeneity} = 0.64) for the pooled results from six studies, while no significant association was found between obesity and risk of early stage (RR 0.91, 95% CI: 0.74–1.08, P = 0.67; I² = 2.5%, P_{heterogeneity} = 0.39) in the pooled analysis of four studies. 

Figure 3

 

Forest plot on the associations between obesity and age-related macular degeneration. The boxes and lines indicate the relative risks (RRs) and their confidence intervals (CIs) on a log scale for each study. The pooled relative risk is represented by a diamond. The size of the black squares indicates the relative weight of each estimate.

These participant characteristics did not significantly alter the shape of association between obesity and AMD risk, although population source seemed to be slightly related with the results. Hospital-based studies tended to report a slightly stronger association of obesity with AMD incidence (RR = 2.35, 95% CI: 0.81–2.88) in comparison with values in volunteers (RR = 1.46, 95% CI: 0.95–1.97) and population-based studies (RR = 1.03, 95% CI: 0.81–2.88; Table 2). In sensitivity analyses, exclusion of individual studies revealed that no single study had a particular influence on the overall results. We found no evidence of publication bias, either with visual assessment of funnel plots or with the Egger's and Begg's tests (Egger's test: P = 0.80, Begg's test: P = 0.23). 

Underweight and AMD Subtypes

Four studies^13,14,25,26 investigated the relationship between underweight and AMD incidence. Of the included studies, three and two studies reported results for early stage and late stage, respectively. Findings from present analysis suggest that there is no significant association between underweight and AMD (RR = 0.96, 95% CI 0.74–1.18, P = 0.74; I² = 0.0%, P_{heterogeneity} = 0.81; Fig. 4). Results for early stage (RR = 0.83, 95% CI: 0.54–1.13, P = 0.30; I² = 0.0%, P_{heterogeneity} = 0.92) and late stage of AMD (RR = 1.13, 95% CI: 0.79–1.48, P = 0.19; I² = 0.0%, P_{heterogeneity} = 0.60) were statistically nonsignificant. The summary RR did not significantly change after one study at a time was removed. Egger's and Begg's tests were not suggestive of publication bias (Egger's test: P = 0.69, Begg's test: P = 0.76). 

Figure 4

 

Forest plot on the associations between underweight and age-related macular degeneration. The boxes and lines indicate the relative risks (RRs) and their confidence intervals (CIs) on a log scale for each study. The pooled relative risk is represented by a diamond. The size of the black squares indicates the relative weight of each estimate.

Dose–Response Meta-Analysis

By using restricted cubic spline model, the result showed a linear relationship between BMI and risk of AMD (P_nonlinearity = 0.17). When linear models were fitted, the risk of AMD increased by 2% (RR = 1.02, 95% CI: 1.01–1.04) for each 1 kg/m² increase in BMI (Fig. 5). The subgroup analysis by AMD stages revealed that the RR of AMD risk per 1 kg/m² increase in BMI was 0.99 (95% CI: 0.97–1.02) and 1.04 (95% CI: 1.02–1.05) in early and late stages, respectively. In sensitivity analysis, assigning different values of BMI to the open-ended exposure categories did not substantially modify our results, suggesting a high stability for the current result. 

Figure 5

 

Relative risks (RRs) and the corresponding 95% confidence intervals (CIs) for the dose–response relationship between body mass index and age-related macular degeneration (AMD) risk. The solid line and the dotted lines represent the estimated RRs and their 95% CIs.

Discussion

This meta-analysis summarized the evidence from all available prospective cohort studies to evaluate associations between overweight/obesity and AMD. Our results showed that overweight and obesity had a slight positive association with risk of AMD, and a significant increased risk of late AMD was noted for obese individuals as compared with subjects in the normal range. Additionally, dose–response analysis displayed a potential linear relationship between BMI and risk of AMD, suggesting that keeping normal body weight and avoiding further weight gain may confer potential protection against this disease. 

Previous studies have shown that the secretion of proinflammatory messengers, such as monocyte chemoattractant protein-1 (MCP-1) and tumor necrosis factor-α (TNF-α), was significantly elevated in obese individuals.^29,30 These proinflammatory factors could regulate migration and infiltration of monocyte, disturbing fundamental functions of the RPE, which contributes to typical retinal changes encountered in AMD.³¹ Furthermore, the biological mechanisms whereby obesity increases risk for AMD may also partly be related to effects of increased adiposity on distribution of macular carotenoids. As the major components of macular pigment, xanthophyll carotenoids including lutein and zeaxanthin are fat-soluble pigments and uniquely concentrated at the central fovea.³² Adipose tissue is a major site of macular carotenoid storage due to the partitioning of carotenoids into fat. As body weight increases, more dietary xanthophyll carotenoids would be absorbed into adipocyte with less of these macular carotenoids available to the macula.^33,34 

Several epidemiologic studies had previously examined the association between excess body weight and the risk of AMD. In the Age, Gene/Environment Susceptibility-Reykjavik Study (AGES) cohort of 2868 participants, greater BMI was a statistically significant independent risk factor associated with the development of incident AMD after adjusting for age and sex.³⁵ This result was in accord with findings of the POLA study, which suggested a 2.29- and 1.54-fold increased risk of late AMD and pigmentary abnormalities in obese subjects aged 60 years or older.³⁶ However, these correlations were not confirmed by Munch et al.,³⁷ who failed to find a significant association between excess body weight and AMD risk in a cross-sectional study. Rather, data from the Beaver Dam Offspring Study (BOSS) even showed that obesity tended to have a protective effect on early AMD in American participants.³⁸ Such inconsistencies may partly be attributed to differences in the study design. In contrast with case–control and cross-sectional studies, prospective cohort studies provide stronger evidence for clarifying the direction of the relationship because this study design does not suffer from recall bias and could largely reduce the likelihood of selection bias as well as reverse causation. Moreover, estimated effects for different study designs can be influenced to varying degrees by diverse sources of bias. The combination of different study designs that provide different strength of evidence would result in substantial heterogeneity. Therefore, only cohort studies were included in the present meta-analysis. 

Findings from this meta-analysis showed that excess body weight was slightly associated with the increase risk of AMD, and a significant association of obesity with late AMD rather than early AMD was noted. A relatively weaker effect of overweight/obesity on early-stage AMD could be partially explained by the absence of conspicuous visual impairment in this stage. Patients in early stage are usually asymptomatic and it is not easy to distinguish them only by routine ophthalmic examination.² Thus, the true association of early AMD was more likely to be underestimated. Moreover, our study also found a greater association among obese individuals than among overweight individuals. This finding was also supported by the additional dose–response analysis, which revealed a linear relationship between BMI and risk of AMD. Therefore, body weight control could be considered one of the effective methods to prevent or delay the progression of AMD. 

It should be noted that confounding factors may have influenced our results. Because all the included studies were observational, the results could be subject to residual or unmeasured confounding. The risk estimates used in the present analysis might not be fully adjusted for, and the way in which main effects and covariates were categorized was not always the same. Hence, pooling of effect estimates obtained from different observational studies is a hard challenge. Smoking is an established risk factor for AMD, and many studies have indicated that smokers tend to increase their BMI less than nonsmokers^39–41; however, adjustment for smoking did not affect the association of obesity with AMD in the present meta-analysis. This suggests that the biological mechanisms through which obesity may increase AMD risk are not mediated by the influence of smoking. This result is also supported by several cohort studies.^14,28 

Our findings have important public health significance. Obesity is widely perceived as one of the most important public health challenges of the 21st century. In 2009 and 2010, the prevalence of obesity was 35.5% among adult men and 35.8% among adult women.⁴² The Beaver Dam Eye Study in the United States reported a 3.1% 15-year cumulative incidence for late AMD in adults aged 43 to 86 years.⁴³ The public health implications for the United States are profound: More than 110,000 cases of severe central vision loss per year from late AMD might be avoided if everyone in the middle-aged and old populations could maintain normal body weight throughout life. 

Our study entailed some potential limitations that may affect the interpretation of the results. First, only a few studies examined this association, which limits the power of meta-analysis. The quantitative analyses in our study were based on prospective cohort studies, which tend to be less susceptible to recall and selection bias than retrospective case–control studies. Furthermore, the included studies had adequate follow-up (most studies lasted a decade) and were of good quality. Thus, the results were considered to be robust. Second, a higher level of BMI tends to be associated with other unhealthy behaviors, such as lower levels of physical activity, lower vegetable consumption, and higher alcohol consumption. Although all the included studies adjusted for potential confounding factors, it is also possible that the observed association between BMI and AMD could be underestimated or overestimated due to unmeasured residual confounding. Third, BMI data collected through self-report in the Physicians' Health Study could cause misclassification bias by underestimating the true BMI, which may have led to underestimation or overestimation of the association.¹⁴ However, high validity has been observed for self-reported and measured height and weight.⁴⁴ Thus, this misclassification should not substantially affect the overall qualitative inference. Fourth, different diagnosis methods could have caused misclassification of the outcome. The accuracy of survey questions and visual acuity criteria for detection of early AMD was limited compared with fundus photography. In the present analysis, only the Physicians' Health Study diagnosed AMD relying on questionnaire and medical records, and the objective of this study was to examine relationships of BMI with visually significant late AMD, rather than early AMD. Since it is less likely that participants with decreased vision would fail to seek medical attention at this stage, the studies pooled in this paper have similar power for late AMD, indicating that the results from the present meta-analysis were reliable. However, such bias cannot be ruled out completely and might have led to underestimation or overestimation of the association. Fifth, the potential for publication bias is of concern, for studies with results that are not statistically significant may be less likely to be published, and the small number of included studies limited the power of Begg's and Egger's tests. Despite the fact that statistical tests did not provide evidence of publication bias in our meta-analysis, it is still hard to fully rule out such bias. Finally, all studies were conducted in white populations in Western countries. Thus, the present results may not be generalizable to other populations. More research needs to be carried out in different countries to examine variations between populations. 

In summary, the results of this meta-analysis demonstrated that excess body weight may act as a potential risk factor for AMD incidence in a dose-dependent fashion, especially with regard to its late stage. With the increasing prevalence of obesity, it is important to realize that reduction in risk of AMD may be an additional benefit of body weight control, especially in obese patients. It should also be noted that only a few prospective studies have been conducted to examine this association, and intervention studies are absent. Therefore, further large-scale long-term prospective cohort and intervention studies investigating the relationship between excess body weight and AMD risk need to be conducted before definitive conclusions can be made. 

Acknowledgments

Supported in part by grants from the National Natural Science Foundation of China (NSFC-81202198, NSFC-81473059); the Natural Science Foundation of Shaanxi Province of China (2013JQ4008); the China Postdoctoral Science Special Foundation (2015T81036); and the China Postdoctoral Science Foundation Funded Project (2014M560790). The authors alone are responsible for the content and writing of the paper. 

Disclosure: Q.-Y. Zhang, None; L.-J. Tie, None; S.-S. Wu, None; P.-L. Lv, None; H.-W. Huang, None; W.-Q. Wang, None; H. Wang, None; L. Ma, None