Hypertensive retinopathy and cardiovascular disease risk: 6 population-based cohorts meta-analysis

Abstract

Background

The cardiovascular risk associated with different levels of hypertensive retinopathy, including mild, remains unclear. We performed an individual participant meta-analysis from 6 population-based cohort studies to determine the relationship of hypertensive retinopathy with incident cardiovascular outcomes.

Methods

We identified cohort studies that objectively assessed hypertensive retinopathy from photographs, documented incident cardiovascular outcomes, and were population-based. Six studies contributed data from 11,013 individuals at baseline with 5–13 years follow-up. Participants were recruited if they had hypertension and did not have confounding conditions such as diabetic retinopathy. Main outcome measures were incident coronary heart disease (CHD), stroke and a composite endpoint of cardiovascular disease (CHD or stroke). Pooled estimates of incident risk ratios (IRR) were obtained after adjusting for age, gender, systolic blood pressure, serum total cholesterol, high density lipoprotein and smoking.

Results

Among eligible participants with hypertension and without diabetes, there were 1018/9662 (10.5%) incident CHD events, 708/11,013 (6.4%) incident stroke events and 1317/9378 (14.0%) incident CVD events. Mild hypertensive retinopathy was associated with increased risk of CVD (IRR 1.13, 95% CI 1.00 to 1.27) and CHD (IRR 1.17, 95% CI 1.02 to 1.34) but not stroke; moderate hypertensive retinopathy was associated with increased risk of CVD (IRR 1.25 95% CI 1.02 to 1.53) but not stroke or CHD individually.

Conclusions

In persons with hypertension, both mild and moderate hypertensive retinopathy were associated with higher CVD risk.

1. Introduction

Hypertension is responsible for one in eight deaths worldwide, and is one of the three leading cause of mortality globally [1]. Hypertensive retinopathy refers to changes in the retinal microvasculature that occur with elevated blood pressure and are a visible manifestation of vascular damage. There are three main stages of hypertensive retinopathy: vasoconstrictive which is characterised by generalised retinal arteriolar narrowing, sclerotic which has focal arteriolar narrowing, arteriolar wall opacification and compression of the venules, and lastly the exudative phase which sees the retinal microaneurysms, haemorrhages, hard exudates and cotton-wool spots appear. Amongst non-diabetic persons, hypertensive retinopathy has been reported to impact 4%–18.7% of the general population, with more men than women impacted by this pathology. Prevalence also varies by ethnicities with Chinese, African Americans, and African Caribbean populations having higher prevalence than white populations [2]. Many national guidelines currently recommend, but do not mandate, ocular fundus examinations for persons with moderate to more severe hypertension to detect presence and severity of hypertensive retinopathy [[3], [4], [5], [6]]. Currently, these guidelines consider only hypertensive retinopathy of moderate or more severe levels to reflect target end organ damage and thus confer increased cardiovascular risk. Hypertensive retinopathy, however, is a progressive medical condition. It is therefore worrying that there is limited information on the risk associated with mild levels of hypertensive retinopathy which would impact subsequent disease treatment and management plans.

To address this important gap, we aimed to determine the risk associated with hypertensive retinopathy in the general population to facilitate wide external applicability. We conducted an individual participant data meta-analysis of population-based studies to determine at what level of hypertensive retinopathy does increased risk of cardiovascular outcomes occur. The outcomes we examined are risk of incident coronary heart disease (CHD), stroke and a combined endpoint of cardiovascular disease (CVD) in persons with hypertension.

2. Methods

2.1. Data sources and inclusion criteria

Study inclusion was decided a priori and stipulated that all studies should have: (1) recruited participants from the general population (i.e. population-based as opposed to clinic- or hospital-based); (2) documented the diagnosis of hypertension and presence of hypertensive retinopathy lesions in an objective, blinded, standardised manner from photographic records (digital or film) that could be used to construct classification grades; (3) were prospective with recorded CVD event data, and; (4) had existing collaboration agreements permitting individual participant data sharing. The population of focus was any participant diagnosed with hypertension. We excluded those with either a history of cardiovascular disease, diabetic retinopathy or with diabetes as this is a confounding variable.

Six studies were identified that met all inclusion criteria. Investigators of these studies were invited to join the collaboration, and provided both published and unpublished data for the individual participant meta-analysis. The six studies were the Atherosclerosis Risk in Communities Study (ARIC) [[7], [8], [9]], the Cardiovascular Health Study (CHS) [10], the Multi-Ethnic Study of Atherosclerosis (MESA) [11], the Beaver Dam Eye Study (BDES) [ [12,13]], (all in the USA) the Blue Mountains Eye Study (BMES) [[13], [14], [15]], (Australia) and the Rotterdam Study (RS) [16] (Europe). The present study was approved by the University of Sydney Human Research Ethics Committee (HREC2016/068), and all participants had provided signed informed consent in the primary studies for their anonymised data to be used in publications. The studies were conducted before patient and public involvement in research became standard.

2.2. Assessment and definition of hypertensive retinopathy

We classified hypertensive retinopathy according to a simplified classification system developed by the authors ourselves following a comprehensive review of the literature [ [17,18]]. The system has been shown to be simple, reproducible and reliable [19]. The simplified hypertensive retinopathy classification includes four categories of increasing severity: none, mild (arteriolar narrowing, arteriovenous nipping or focal arteriolar narrowing), moderate (microaneurysms, haemorrhages, hard or soft exudates), and severe (optic disc swelling). Supplementary Fig. 1 summarises these grades and shows examples. Individual retinopathy lesions were assessed from retinal photographs or digitised slides in all studies by trained graders blinded to participant age, blood pressure and other information. Average arteriolar and venular calibres at pre-specified points away from the optic disc were graded using computer-assisted software [17]. Inter-grader reliability statistics were good, with intra-class correlation coefficients of 0.71–0.99 in the different studies.

Participants in all studies completed baseline questionnaires including previous medical history. All studies used standard methods to measure Framingham CVD risk factors. Diabetes was defined according to a combination of fasting or random blood glucose levels, a physician diagnosis or current use of diabetes medication. For hypertension, all studies collected similar data and we were able to standardise a definition as systolic blood pressure ≥140 mmHg and/or use of blood pressure-lowering medication.

2.3. Outcomes of interest

The primary outcome was CVD events with stroke and CHD taken as secondary outcomes. Briefly, CHD and stroke event data were obtained as either fatal or non-fatal. These were combined to form the endpoints CHD, stroke and the composite CVD (CHD or stroke). Non-fatal CHD events were defined as angina episodes, myocardial infarction, coronary artery bypass graft, and coronary angioplasty as reported by the studies. Non-fatal stroke events excluded transient ischemic attacks. Table 1 summarises how each CHD and stroke events were measured by each included study.

Table 1.

Assessment criteria of fatal and non-fatal CHD and stroke events in each included study.

Study	Outcomes of Interest
	CHD and Stroke Events
ARIC	Fatal: Death registries, death certificates, and coroner's reports Non-fatal: annual phone contact with participants, followed by contact with physicians to help classify events once identified.
BDES	Fatal: Death registries and at 5-year repeat examinations Non-fatal: participants returning to subsequent visits were asked whether they had CHD or stroke event.
BMES	Fatal: Death registries and at 5-year repeat examinations Non-fatal: participants returning to subsequent visits were asked whether they had CHD or stroke event with answers verified from medical records.
CHS	Fatal: Death registries, death certificates, and coroner's reports Non-fatal: continuous monitoring including regular phone interviews with participants and contact with general practitioners and local hospitals.
MESA	Fatal: Death registries, death certificates, and coroner's reports Non-fatal: continuous monitoring including regular phone interviews with participants and contact with general practitioners and local hospitals.
Rotterdam Study	Fatal: data linkage with general practitioner and municipality databases Non-fatal: data linkage with general practitioner and municipality databases and review of nursing home records.

2.4. Statistical analysis

We used multiple Poisson regression models to estimate incidence rate ratios (IRR) which are similar to relative risks averaged over the follow-up period. This method has advantages when analysing rare events in individual patient data meta-analyses, particularly when studies have different follow-up times [20]. We confirmed these analyses using the Kaplan-Meier method and Cox proportional hazard models to estimate hazards ratios.

Separate fixed effects meta-analyses for each outcome (CHD, stroke and CVD) were conducted using two-stage individual patient data meta-analysis models. Individual patient data were first used in each study separately, and in the second stage the summary data (IRR) were synthesised using a traditional model for meta-analysis of aggregate data. We adjusted for gender, age, ethnicity, systolic blood pressure, serum total cholesterol, high density lipoprotein, smoking status and use of blood pressure lowering medication in the multivariable Poisson regression models for survival analysis.

We assessed heterogeneity between studies with the I² statistic as a measure of the proportion of total variation in estimates due to heterogeneity. P-values <0.05 were considered statistically significant. Possible sources of publication bias were assessed with contour-enhanced funnel plots and formal statistical tests. All statistical analyses were performed using Stata 12.1 (StataCorp, College Station, Texas). This study complies with both MOOSE and STROBE guidelines [ [21,22]].

3. Results

This meta-analysis comprised data from six studies. There were 28,275 participants with retinal and CVD data. We excluded 12,683 participants without hypertension, 2642 with diabetes, and 3572 with a history of CVD. This left 9378 participants who were included in analyses for incident CVD outcomes (9662 participants for incident CHD, and 11,013 participants for incident stroke). Prevalence of mild hypertensive retinopathy in the studies ranged from 22.5% (RS) to 47.2% (ARIC) with an aggregate prevalence of 33.7%; for moderate retinopathy, prevalence ranged from 4.5% (ARIC) to 11.3% (MESA) with an aggregate prevalence of 7.8%. No participant in any study was recorded as having severe hypertensive retinopathy. Table 2 shows baseline characteristics of participants at risk of incident CVD. Follow-up ranged from a median of 4.8 years (BMES, MESA) to 13.0 years (BDES), over which period we recorded 1018 (10.5%) incident CHD, 708 (6.4%) incident stroke and 1317 (14.0%) incident CVD events.

Table 2.

Baseline characteristics of participants with hypertension^a included in meta-analysis for cardiovascular disease analyses (n = 9378).

Study	Participants at risk of CVD, N	Simplified Hypertensive Retinopathy Grade			Participant Characteristics
		None n (%)	Mild n (%)	Moderate n (%)	Follow-up, median years	Age, mean (SD) years	Female, n (%)	SBP, mean (SD), mmHg	Serum cholesterol, mean (SD), mmol/L	Serum HDL, mean (SD), mmol/L	Current smoker, n (%)
ARIC	2811	1412 (50.2)	1270 (45.2)	129 (4.6)	9.1	60.5 (5.6)	1675 (59.6)	137.1 (19.2)	5.4 (0.9)	1.4 (0.5)	425 (15.1)
BDES	1256	745 (59.3)	390 (31.1)	121 (9.6)	13.0	62.4 (10.3)	752 (59.9)	142.4 (19.0)	6.2 (1.1)	1.4 (0.5)	206 (16.4)
BMES	831	484 (58.2)	255 (30.7)	92 (11.1)	4.8	65.1 (8.6)	519 (62.5)	152.9 (16.6)	6.1 (1.0)	1.5 (0.4)	89 (10.7)
CHS	597	343 (57.5)	221 (37.0)	33 (5.5)	6.4	78.6 (4.1)	396 (66.3)	138.4 (20.4)	5.4 (1.0)	1.3 (0.4)	35 (5.9)
MESA	2071	1246 (60.2)	591 (28.5)	234 (11.3)	4.8	65.4 (9.4)	1149 (55.5)	140.1 (20.4)	5.0 (0.9)	1.3 (0.4)	214 (10.3)
Rotterdam Study	1812	1266 (69.9)	419 (23.1)	127 (7.0)	11.9	68.2 (7.5)	1150 (63.5)	140.1 (20.4)	6.7 (1.2)	1.4 (0.4)	326 (18.0)
All studies combined	9378	5496 (58.6)	3146 (33.6)	736 (7.9)	7.8	64.9 (9.1)	5641 (60.2)	142.5 (20.0)	5.7 (1.2)	1.4 (0.4)	1295 (13.8)

SD, standard deviation; SBP, systolic blood pressures; HDL, high density lipoprotein; ARIC, Atherosclerosis Risk in Communities Study; BDES, Beaver Dam Eye Study; BMES, Blue Mountains Eye Study; CHS, Cardiovascular Health Study; MESA, Multi-Ethnic Study of Atherosclerosis.

^aHypertension defined as systolic blood pressure ≥140 mmHg and/or using blood pressure lowering medication.

3.1. Individual studies results

Mild hypertensive retinopathy was associated with incident CVD only in CHS (IRR 1.74, 95% confidence interval (CI) 1.26 to 2.40), whereas moderate hypertensive retinopathy was associated with incident CVD only in ARIC (IRR 1.59, 95% CI 1.02 to 2.46) (Fig. 1 and Supplemental Table 1). Mild hypertensive retinopathy was not associated with incident CHD in any individual study (data not shown). Similarly, moderate hypertensive retinopathy was not associated with CHD in individual studies, except ARIC. For stroke, a similar pattern was observed, with mild retinopathy not associated with stroke in any individual study. However, moderate retinopathy was associated with incident stroke in ARIC, BMES and MESA (data not shown). All other studies showed a direction of association favouring an association with stroke, except BDES.

Fig. 1.

Forest plots showing association of mild (A) and moderate (B) hypertensive retinopathy graded according to the simplified classification of hypertensive retinopathy and risk of cardiovascular disease (CVD) in people with hypertension (n = 9378). Analyses are adjusted for age, gender, ethnicity, systolic blood pressure, serum total cholesterol, high density lipoprotein and smoking status.

IRR, incidence rate ratios; CI, confidence interval; CVD, cardiovascular disease (defined as coronary heart disease or stroke); ARIC, Atherosclerosis Risk in Communities Study; BDES, Beaver Dam Eye Study; BMES, Blue Mountains Eye Study; CHS, Cardiovascular Health Study; MESA, Multi-Ethnic Study of Atherosclerosis; Rotterdam, Rotterdam Study.

3.2. Pooled results

When all studies were combined, both mild and moderate hypertensive retinopathy were associated with incident CVD (IRR 1.13, 95% CI 1.00 to 1.27; IRR 1.25, 95% CI 1.02 to 1.53, respectively) (Fig. 1 and Table 3). These associations were similar in men and women, although did not reach statistical significance in these subgroups. There was no statistically significant interaction of gender on IRRs for CVD, stroke or CHD.

Table 3.

Simplified classification of hypertensive retinopathy and risk of coronary heart disease (CHD), stroke and cardiovascular disease (CVD) in participants with hypertension^b, meta-analysis of six population-based studies.

All studies combined	Simplified Hypertensive Retinopathy Grade
	None			Mild			Moderate
	N at risk	% CVD events	IRRa (95% CI)	N at risk	% CVD events	IRRa (95% CI)	N at risk	% CVD events	IRRa (95% CI)
All
CVD	5496	13.41	1.00	3146	14.88	1.13 (1.00, 1.27)	736	15.22	1.25 (1.02, 1.53)
Stroke	6550	6.50	1.00	3588	5.74	0.93 (0.78, 1.11)	875	8.69	1.50 (0.88, 2.56)
CHD	5629	9.66	1.00	3257	11.91	1.17 (1.02, 1.34)	776	11.08	1.15 (0.91, 1.45)
Men
CVD	2100	17.33	1.00	1329	19.04	1.11 (0.94, 1.32)	308	18.51	1.25 (0.94, 1.66)
Stroke	2667	6.94	1.00	1557	5.96	0.92 (0.71, 1.21)	376	7.45	1.31 (0.87, 1.98)
CHD	2126	13.75	1.00	1375	15.93	1.13 (0.94, 1.37)	325	15.69	1.29 (0.95, 1.75)
Women
CVD	3396	10.98	1.00	1817	11.83	1.13 (0.95, 1.34)	428	12.85	1.25 (0.94, 1.67)
Stroke	3883	6.21	1.00	2031	5.57	0.96 (0.76, 1.22)	499	17.12	1.67 (0.90, 3.12)
CHD	3503	7.12	1.00	1882	8.98	1.16 (0.95, 1.43)	451	7.76	1.13 (0.79, 1.63)

Bold denotes significanceIRR, incidence rate ratio; CI, confidence interval; CVD, cardiovascular disease (defined as coronary heart disease or stroke).

^aAdjusted for age, gender, ethnicity, systolic blood pressure, serum total cholesterol, high density lipoprotein and smoking status.

^bHypertension defined as systolic blood pressure ≥140 mmHg and/or using blood pressure lowering medication.

For CHD and stroke outcomes separately, the pooled estimate for mild hypertensive retinopathy and incident CHD was significant (IRR 1.17, 95% CI 1.02 to 1.34), but moderate retinopathy and incident CHD was not (IRR 1.15, 95% CI 0.91 to 1.45) (Table 3). Pooled estimates for mild and moderate hypertensive retinopathy and incident stroke were not significant (IRR 0.93, 95% CI 0.78 to 1.11; IRR 1.50, 95% CI 0.88 to 2.56, respectively).

In all fixed-effects meta-analyses, the I² statistic showed no evidence of heterogeneity. Results using Cox proportional hazard models were similar and confirmed findings using Poisson regression models (Supplemental Table 2). Using different definitions of arteriolar narrowing for mild retinopathy resulted in very similar risk estimates. When arteriolar narrowing was defined as the smallest quintile of AVR, mild retinopathy and moderate retinopathy associations with CVD outcomes were IRR 1.25 (95% CI 1.11 to 1.40) and IRR 1.31 (95% CI 1.07 to 1.60), respectively. When arteriolar narrowing was defined as either the narrowest quintile of arteriolar calibre or widest quintile of venular calibre, the corresponding IRR for mild and moderate retinopathy and CVD outcomes were 1.14 (95% CI 1.02 to 1.28) and 1.28 (95% CI 1.04 to 1.57), respectively.

Funnel plots for possible sources of publication bias did not show asymmetry or substantial evidence of publication bias, and both Begg's and Egger's regression asymmetry tests showed no evidence of substantial publication bias.

4. Discussion

We report that in participants with hypertension and free from diabetes, mild hypertensive retinopathy was associated with increased risk of incident CVD and CHD events over 5–13 years. Moderate retinopathy was associated with a similar but slightly higher risk of incident CVD events. These associations were largely consistent across all six studies and similar in men and women.

Our results provide evidence that even mild levels of hypertensive retinopathy are associated with increased risk of CVD outcomes. The presence of such signs may be an indicator of individual susceptibility to CVD from elevated blood pressure. We previously reported in a study of 1120 persons who underwent coronary angiography that those with narrower retinal arterioles (mild hypertensive retinopathy) were more likely to have significant coronary artery stenosis [23]. Tedeschi-Reiner and colleagues reported in a study of 109 patients who underwent coronary angiography that more severe retinal arteriolar narrowing was associated with more severe coronary artery disease [24]. Our group and others have also found that patients with acute ischemic stroke are more likely to have mild and moderate hypertensive retinopathy signs, even after adjusting for age, blood pressure and other risk factors [ [25,26]]. These results support the concept that microvascular changes in retinal vessels may reflect vascular disease in the coronary and cerebral circulations [27].

Our risk estimates for mild hypertensive retinopathy are similar, but slightly lower than those from other studies using subjective evaluation of mild hypertensive retinopathy changes. The Ibaraki Prefectural Health Study [28] subjectively examined earlier stages of hypertensive retinopathy according to the older Keith-Wagner-Barker (KWB) classification [ [29,30]] and reported multivariable hazard ratios for total CVD mortality of 1.24 (95% CI 1.12 to 1.38) for grade 1 (generalised arteriolar narrowing only) and 1.23 (95% CI 1.03 to 1.47) for grade 2 (focal narrowing, arteriovenous nipping) among men, and 1.12 (95% CI 1.01 to 1.24) and 1.44 (95% CI 1.24 to 1.68) for grades 1 and 2, respectively, among women [28]. Both these grades combine into mild hypertensive retinopathy on the simplified scale, and the risk estimates are similar to those in our study. Similarly, a 2022 Japanese study also using the KWB classification found mild hypertensive retinopathy to be associated with risk of CVD (HR: 1.34; 95% Cl 1.04 to 1.49) and stroke (HR: 1.28 95% Cl 1.01 to 1.62) [31]. The First National Health and Nutrition Survey (NHANES I) [32] subjectively assessed early hypertensive retinopathy on ophthalmoscopy (equivalent to mild retinopathy on the simplified scale) and reported a multivariable relative risk for incident CVD of 1.2 (95% CI 1.0 to 1.3), which is also similar to the estimate from our study. However, these studies were limited in that they assessed hypertensive retinopathy subjectively, assessors were not blinded to participants’ blood pressure readings, they included participants with the major confounding factors of diabetes and diabetic retinopathy, and they did not assess moderate hypertensive retinopathy. Guidelines for risk calculation often automatically assign a high risk to persons with diabetes, so they do not require further risk stratification and may confound studies of hypertensive retinopathy [3]. Further, both diabetes [33] and diabetic retinopathy [34] are known to increase cardiovascular risk independently of hypertensive retinopathy changes. As far as we are aware, our study is the first to report on both mild and moderate hypertensive retinopathy using meta-analysis in the same study, and the first to exclude participants with diabetes and diabetic retinopathy. These exclusions may account for the slightly lower CVD risk we report in our meta-analysis.

Our findings add to accumulating evidence from the Ibaraki Prefectural Health Study and others [28,32], that mild hypertensive retinopathy may be associated with increased cardiovascular risk and might be an indicator of target end organ damage. This may have implications towards future guidelines on how hypertension is managed. It would however be premature to consider changing guidelines at this stage as more evidence would be needed to show that interventions at this early stage are beneficial. A review of 19 clinical trials on antihypertensive drugs found all trials reported statistically significant lowering of blood pressure and risk of cardiovascular disease but not mortality [35]. However, not all of these studies included participants with mild hypertension or hypertensive retinopathy. Results from the Australian National Blood Pressure Study, which did look at the efficacy of blood pressure-lowering drug treatment in mild hypertensive people, found pharmaceutical treatment for cardiovascular disease based upon the person's risk of disease to be better than plans based off their blood pressure thresholds alone [36]. These results are also supported by a meta-analysis of 47,872 participants which showed drug treatment plans based upon cardiovascular risk to prevent more cardiovascular events than those based upon a wide range of blood pressure thresholds [37]. With our study showing mild and moderate hypertensive retinopathy increases the risk of CVD, future trials focusing on this subset of patients may be useful in collecting more evidence for future updates to treatment and management guidelines. The cost effectiveness of screening with retinal photography or eye exams should also be considered when evaluating the clinical utility of assessing for hypertensive retinopathy.

The strengths of this meta-analysis include access to individual participant data which enabled us to standardise many factors including inclusion criteria, grading of hypertensive retinopathy, adjustment for confounders and definitions of CVD events. Hypertensive retinopathy signs were assessed by graders blinded to blood pressure and other participant information in an objective, similar and reproducible manner in all studies. In several studies the same graders assessed the retinal photographs. Access to individual participant-level data allowed us to calculate more accurate risk estimates than meta-analyses which rely on published estimates. Several limitations deserve mention. First, it is possible that participants who had hypertensive retinopathy changes at baseline were treated more aggressively, resulting in a reduced rate of CVD events. We are unable to adjust for this as not all studies collected data on repeated measures of blood pressure or changes in anti-hypertensive medication use. This effect may act to reduce the risk estimate we report. We were not able to adjust for other potential confounders such as body mass index, high-sensitivity C-reactive protein levels or carotid artery disease, as these were not routinely collected in all studies. In the ARIC study, the proportion of mild retinopathy was higher, while that of moderate retinopathy was lower, than other studies, This may be related to the younger age of the ARIC cohort. While ARIC showed significance for moderate retinopathy, all the other studies except BDES also showed a possible effect in the same direction. For mild retinopathy, the results were mainly driving by CHS and to a lesser extent, ARIC. This may be because the CHS and ARIC were specifically designed to detect CVD outcomes, whereas most of the other studies were not.

In conclusion, this individual participant meta-analysis found both mild and moderate hypertensive retinopathy are independent risk factors for CVD events. Although our findings add to evidence that mild hypertensive retinopathy may be associated with slightly increased risk, the clinical implications of these findings at present remain unclear. Further work is needed as to whether interventions to further reduce blood pressure or CVD risk factors in patients with mild hypertensive retinopathy results in improved CVD outcomes.

Credit author statement

Gerald Liew: Conceptualisation, Methodology, Write-up of original manuscript; Jing Xie: Conceptualisation, Methodology, Analysis, Write-up of original manuscript; Helen Nguyen: Methodology, Write-up of original manuscript; Lisa Keay: Methodology, Revision of original manuscript; M Kamran Ikram: Conceptualisation, Revision of original manuscript; Kevin McGeechan: Conceptualisation, Revision of original manuscript; Barbara EK Klein: Conceptualisation, Revision of original manuscript; Jie Jin Wang: Conceptualisation, Methodology, Revision of original manuscript; Paul Mitchell: Conceptualisation, Methodology, Revision of original manuscript; Caroline CW KLaver: Conceptualisation, Methodology, Revision of original manuscript; Ecosse L Lamoureux: Conceptualisation, Revision of original manuscript; Tien Y Wong: Conceptualisation, Methodology, Revision of original manuscript.

Funding

The Cardiovascular Health Study is supported by contracts N01-HC-85079 through N01-HC-85086, N01-HC-35129, N01 HC-15103, N01 HC-55222, N01-HC-75150, N01-HC-45133, grant number U01 HL080295 from the National Heart, Lung, and Blood Institute, with additional contribution from the National Institute of Neurological Disorders and Stroke. A full list of principal CHS investigators and institutions can be found at http://www.chs-nhlbi.org/pi.

The Atherosclerosis Risk in Communities Study is carried out as a collaborative study supported by National Heart, Lung, and Blood Institute contracts (HHSN268201100005C, HHSN268201100006C, HHSN268201100007C, HSN268201100008C, HSN268201100009C, HHSN268201100010C, HSN268201100011C, and HHSN268201100012C). The authors thank the staff and participants of the ARIC study for their important contributions.

The Multi-Ethnic Study of Atherosclerosis study is supported contracts N01-HC-95159 through N01-HC-95166 from the National Heart, Lung, and Blood Institute. Additional support was provided by NIH grant HL69979-03 (Klein R and Wong TY).

The Blue Mountains Eye Study was supported by the Australian National Health & Medical Research Council, Canberra, Australia (NHMRC Grant No:153,948, 302068, 211069).

The Rotterdam study was supported by the Netherlands Organization for Scientific Research (NWO, 91203014, 175.010.2005.011, 91103012).

All researchers are independent from the funders listed above. The funders had no role in study design, data collection and analysis and interpretation of data, decision to publish, or preparation of the manuscript.

Declaration of competing interest

The authors report no relationships that could be construed as conflict of interest.

Handling Editor: D Levy

Appendix A. Supplementary data

The following are the Supplementary data to this article.