Temporal variation in cardiovascular disease risk predicted by albuminuria: An opportunity for clinical intervention?

doi:10.4236/ojpm.2013.31003

Paper Menu >>

Journal Menu >>

Vol.3, No.1, 22-27 (2013) Open Journal of Preventive Medicine

http://dx.doi.org/10.4236/ojpm.2013.31003

Temporal variation in cardiovascular disease risk

predicted by albuminuria: An opportunity for clinical

intervention?

Katina D’Onise1, Robyn McDermott1, Adrian Esterman1, Bradley McCulloch2

1Sansom Institute, University of South Australia, Adelaide, Australia; Katina.d’[email protected]

2Tropical Regional Services, Queensland Health and School of Public Health, James Cook University, Townsville, Australia

Received 27 September 2012; revised 31 October 2012; accepted 8 November 2012

ABSTRACT

Albuminuria predict s cardiov ascular disease (CVD)

events but it is likely to vary over time in a non-

linear fashion. The aim of this study was to es-

timate the potentially differing predictive effect

of albuminuria on the risk of CVD or related

death over time. Data were from a cohort study

of 3505 predominately indigenous adults from

remote communities in Queensland, Australia,

1999-2006. Cox Proportional Hazards model

analysis of the predictive effects of urinary al-

bumin creatinine ratio on the risk of CVD or

CVD-related death was undertaken for incident

and prevalent CVD. Analyses sequentially re-

moved those w ho had a cardiovascular event or

related death for the first year through to six

years. The baseline prevalence of microalbu-

minuria was 21.2% and for macroalbuminuria

6.7%. The incidence of CVD was 92 in 13,812

person-years. Microalbuminuria predicted inci-

dent CVD with a Hazard Ratio (HR) of 3.0 (95 % CI

1.83 - 4.96) and for macroalbuminuria HR 10.8

(95% CI 6.58 - 17.68) and for those with pre-ex-

isting CVD, HR 2.6 (95% CI 1.65 - 3.97) and HR

9.7 (95% CI 6.38 - 14.82) respectively. People

with macroalbuminuria who survived the first

three years had a crude HR of an incident car-

diovascular event or death of 13.0 (95% CI 6.45 -

26.39) to a peak of 32.3 (95% CI 8.55 - 121.77) for

those who survived the first five years. The

hazard appeared t o drop in the 6th y ear although

this is based on small numbers. The first three

years after finding macroalbuminuria provide a

potential window opportuni ty to activ ely manage

the risk of incident C VD before the risk elevates .

Keywords: Cardiovascular Diseases; Albuminuria;

Mortality; Risk Factors; Epidemiology

1. INTRODUCTION

Albuminuria is a well established independent risk

factor for cardiovascular disease events and associated

mortality. This has been reported in both high risk and

general population cohorts [1-3]. The mechanism for this

association is not clear, however a number of hypotheses

have been proposed. The association may be accounted

for by albuminuria as a marker of reduced renal function,

which is associated with increased cardiovascular disease

(CVD) risk [4]. Additionally, it may be that widespread

endothelial damage leads both to albuminuria and CVD

through enhanced atherogenesis [4]. It is also possible

that a common factor, such as hypertension, leads to both

albuminuria and CVD [5].

The range of studies reporting this association assume

constant risk (or hazard) over time. This is unlikely to be

true biologically as a high level of macroalbuminuria is

more likely in an advanced stage of disease such that the

risk of a cardiovascular event is greater in the short term

than those with lesser levels of macroalbuminuria, all

other factors equal. As such, using a model such as the

Cox model which assumes constant hazard over time

may not adequately estimate the risk of albuminuria for

cardiovascular disease. In this scenario, the hazard over

the length of the study period may overestimate the risk

in the short term and underestimate it in the long term.

This has implications on the importance clinicians place

on management of albuminuria in the prevention of both

advanced renal and cardiovascular disease in the short

term to reduce greater risk in the long term.

To examine this further, we used data from the Well

Person’s Health Check (WPHC), a cohort study in North

Queensland conducted in 26 remote Aboriginal and Tor-

res Strait Islander communities [6]. This general popula-

tion cohort has previously documented high levels of

albuminuria and CVD [7] and so provides an opportunity

to examine the effect of changing risk of albuminuria on

cardiovascular disease over time.

K. D’Onise et al. / Open Journal of Preventive Medicine 3 (2013) 22-27 23

2. MATERIALS AND METHODS

2.1. Study Population

Baseline data were collected from 3505 people in 26

rural Indigenous communities in Far North Queensland,

who participated in the “Well Person’s Health Check”

between 1999 and 2000 (methods for this cross-sectional

study have been reported in detail elsewhere [6]). Based

on the local census data, the study achieved a participa-

tion rate of 44.5% with greater participation noted in

smaller communities.

2.2. Albumin Creatinine Ratio

All participants were asked to provide urine from the

first morning void or to delay providing the sample for at

least two hours from the most recent void. First catch

urine samples were self-collected in a sterile 50 ml con-

tainer. All urine specimens were refrigerated at four to

eight degrees centigrade immediately following collec-

tion. A five to ten millilitre sample from the collection

jar was transferred into a ten-millilitre tube. Dipstick

urinalysis (Combur-test, Roche) was performed on the

remaining sample and the results for protein recorded.

For the first 15 months of the recruitment period (of a

total recruitment period of 34 months) albumin creatinine

ratio (ACR) testing was performed when the participants’

urine contained protein (detected on urinalysis) or they

were known to have diabetes, hypertension or had a body

mass index over 30 kg/m2. For the subsequent 19 months

of recruitment ACR testing was performed routinely on

all urine specimens [6]. Microalbuminuria was defined

as a urinary ACR of 3.4 mg/mmol to 33.9 mg/mmol and

macroalbuminuria of 34 mg/mmol or greater.

2.3. CVD Event or Related Death

A cardiovascular event or related death was deter-

mined from unit-level record linkage of the cohort mem-

bers to hospitalisation records and deaths up to a census

date of 1 January 2006. Hospitalisation and death records

for consenting WPHC participants were identified by a

manual search (conducted by a registered nurse with

experience working in the region) of the Queensland

Health hospital records systems. As there is no unique

patient identifier in Queensland, a mapping table, which

linked WPHC reference number, hospital facility code

and local unit record number, was developed. This table

was subsequently applied to the Queensland Hospital

Admitted Patient Data Collection, and hospitalisation

relevant to the match unit record, facility code tuples

were extracted. Matching of death records was per-

formed manually at the Queensland Registry of Births,

Deaths and Marriages.

Hospitalisations were considered to be CVD related if

they contained an International Classification of Diseases,

ninth revision, clinical modification (ICD-9-CM) code

commencing with 410, 411, 413 or 414, or an ICD-9-CM

procedure code between 3600 and 3699, inclusive. For

hospitalisations coded to the International Classification

of Diseases, 10th revision, the diagnosis code range I20 -

I25, and procedure code blocks 669 - 679, inclusive,

were used.

Both incident CVD and a separate analysis on the

whole sample that also included those with pre-existing

CVD were undertaken. Prior cardiovascular disease was

determined from participant report of a history of car-

diovascular disease or from a case note review at study

entry or hospitalisation (using the same International

Classification of Diseases codes as above) recorded prior

to study entry.

2.4. Baseline Data

Current smoking was defined as regular smoking of at

least one cigarette a day. Blood pressure was the average

of three measurements taken while seated over a ten

minute period. Height and weight were measured, with

weight recorded to the nearest 0.1 kg and height recorded

to the nearest centimetre. Body mass index was calcu-

lated as weight (kg) divided by the height squared (m2)

and categorised by using WHO definition of overweight

and obesity [8].

Blood for total cholesterol was collected by trained

health staff. Blood was collected in an eight point millili-

tre clotted (SST) vacuum tube which was spun for 10

minutes in a portable centrifuge within one hour of col-

lection and measured by photometric enzyme end point

assay, using the Cobas Integra 700/400 (Roche Diagnos-

tics, New York, USA).

Prior history of diabetes at baseline was determined

from a report at interview of diabetes or case note review

or a fasting blood glucose level of ≥7 mmol/L, 2-hour

glucose tolerance test result of ≥11.1 mmol/L at study

entry, or documented hospitalisation for diabetes prior to

study entry (through the linkage process described above

for ICD-09 codes E11-E14).

2.5. Study Sample

Those who did not have an ACR measured (n = 643)

and missing data on variables blood pressure or smoking

(a further n = 413), date of the first screen (n = 1) or for

cause of death (n = 3) lead to a sample for analysis of

2445. Exclusion of those with pre-existing CVD brought

the sample to 2349. As noted above, the majority of the

ACR data were missing due to the selection criteria of

the study and not at random.

2.6. Analysis

A descriptive analysis was undertaken comparing

K. D’Onise et al. / Open Journal of Preventive Medicine 3 (2013) 22-27

those participants who did and did not have an ACR

taken or had missing data was undertaken, using chi

square test or a student’s t-test.

All CVD related deaths or events (CVD) were pooled

for subsequent analyses. Time in the study was calcu-

lated from date of first screen to CVD or death from any

cause or the census date (whichever occurred first). Car-

diovascular event or death rates per 1000 person-years

were analysed using the Kaplan-Meier method by ACR

category.

The effect of ACR level at baseline on CVD was ana-

lysed using a Cox Proportional Hazards model, estimat-

ing a Hazard Ratio (HR). This analysis was conducted on

the entire study sample with a total of up to seven years

of follow up (n = 2445), assuming proportional hazards

over time. In order to assess the potential effect of dif-

fering hazard over time, the analysis was repeated by

excluding all those events or deaths that occurred in the

first year subsequent to study entry and then again ex-

cluding the events or deaths that occurred in the first two

years subsequent to study entry, and so on until six years

from study entry. In this way these models assess the

subsequent risk of a CVD event or related death for peo-

ple who survived from one to six years following study

entry.

Crude Hazard Ratios were calculated using Cox mod-

els. The independent contribution of albuminuria as a

marker of cardiovascular disease risk was estimated us-

ing multivariate analysis. For each of the time periods,

the analysis was conducted first adjusted for sex and age

only (model 2, there was minimal difference in estimates

stratified by gender and so to maximize precision the

analysis controlled for gender rather than stratifying by

gender). The second model further adjusted for a history

of prior CVD and the third model also adjusted for

smoking status, systolic blood pressure (adding diastolic

blood pressure to the model did not appreciably alter

results and so was omitted, data not shown), diabetes,

body mass index and total cholesterol level, factors that

have been previously found to have an association both

with CVD and ACR [4,9,10]. All analyses were con-

ducted using Stata statistical software v 10.0 [11]. The

study was approved by the Cairns and Hinterland Health

Service District Ethics Committee, with support from the

relevant peak Indigenous health councils. All participants

gave informed consent.

3. RESULTS

The study cohort had a 21.2% baseline prevalence of

microalbuminuria and 6.7% of macroalbuminuria, with

no difference between men and women (p = 0.3). Table

1 shows the baseline characteristics of the study popula-

tion, comparing the study sample with those who par-

ticipated in the baseline survey but either did not have an

ACR taken (as they were deemed to be ‘low risk’) or had

missing data. The study sample, was more likely to be

female, had higher blood pressure, higher BMI, lower

total cholesterol, higher prevalence of diabetes and was

older, but there was no difference in the prevalence of

smoking.

The rate of incident CVD events or death per 1000

person-years by ACR category and length of time since

study entry is presented in Table 2. The overall inci-

dence of cardiovascular event or death was 92 in 13,812

person-years. The rate increased across ACR category

from normal (3.3, 95% CI 2.3 - 4.6) to microalbuminuria

(9.8, 95% CI 6.8 - 14.1) and then macroalbuminuria

(35.5, 95% CI 24.8 - 50.8).

Kaplan-Meier survival estimates are presented in the

Figure 1. There was decreasing survival with increases

in level of albuminuria.

The overall risk of an incident cardiovascular event or

death was 1.3 (95% CI 0.78 - 2.31) for microalbuminuria

Table 1. Baseline characteristics of the study sample (analysis cohort) and those who had missing data and were excluded from the

analysis, Well Person’s Health Check, 1999-2006, n = 3377.

Variable (mean unless otherwise indicated) Study sample n = 2445

(95% CI†)

Missing ACR and other covariates n = 932

(95% CI)

Age* 36.9 (36.2 - 37.5) 34.2 (33.3 - 35.1)

Female* (%) 53.2 (51.2 - 55.1) 47.8 (44.7 - 50.7)

Smoker (%) 51.8 (49.8 - 53.8) 52.2 (49.2 - 55.3)

Systolic blood pressure* 129.5 (128.7 - 130.2) 124.6 (123.5 - 125.6)

Body Mass Index* 27.7 (27.4 - 28.0) 26.1 (25.7 - 26.5)

Total cholesterol* 4.8 (4.78 - 4.87) 5.0 (4.9 - 5.1)

Diabetes*(%) 14.3 (12.9 - 15.7) 8.9 (7.2 - 10.6)

*p value ≤ 0.05; †CI confidence interval.

K. D’Onise et al. / Open Journal of Preventive Medicine 3 (2013) 22-27 25

Table 2. Incident cardiovascular events or death rates per 1000 person years, by category of albuminuria and years since study entry,

Well Person’s Health Check, 1999-2006, n = 2349.

Normal Microalbuminuria Macroalbuminuria

Time

(Years) Person-years EventsRate (95% CI†) Person-yearsEventsRate (95% CI†)Person-years Events Rate (95% CI†)

0 - 1 1687.4 6 3.6 (1.6 - 7.9) 493.9 5 10.1 (4.2 - 24.3)150.8 10 66.3 (35.7 - 123.2)

1 - 2 1675.1 4 2.4 (0.9 - 6.4) 487.9 7 14.3 (6.8 - 30.1)142.6 2 14.0 (3.5 - 56.1)

2 - 3 1664.6 8 4.8 (2.4 - 9.6) 476.3 5 10.5 (4.4 - 25.2)138.2 2 14.5 (3.6 - 57.9)

3 - 4 41654.6 8 4.8 (2.4 - 9.7) 468.5 4 8.5 (3.2 - 22.7)136.6 4 29.3 (11.0 - 78.0)

4 - 5 1644.4 4 2.4 (0.9 - 6.5) 462.0 5 10.8 (4.5 - 26.0)131.5 4 30.4 (11.4 - 81.0)

5 - 6 1123.8 2 1.8 (0.4 - 7.1) 355.4 2 5.6 (1.4 - 22.5)90.8 7 77.1 (36.8 - 161.7)

6 - 7 488.0 1 2.0 (0.3 - 14.5) 186.0 0 0 46.4 1 21.5 (3.0 - 153.0)

7+ 73.2 0 - 27.7 1 36.1 (5.1 - 256.1)7.0 0 0

Total 10011.1 33 3.3 (2.3 - 4.6) 2957.8 29 9.8 (6.8 - 14.1)843.9 30 35.5 (24.8 - 50.8)

†CI: confidence interval.

0.00 0.25 0.50 0.75 1.00

Survival estimate

0 2 4 6 8

Tim e (years)

Normal ……………..

Microalbuminuria

Macroalbuminuria -------------

Figure 1. Kaplan-Meier survival estimate by albuminuria

category for incident cardiovascular disease, Well Person’s

Health Check, 1999-2006, n = 2349.

and 4.4 (95% CI 2.52 - 7.83) for macroalbuminuria

compared with normal levels of albuminuria, when ad-

justing for age, gender, smoking, systolic blood pressure,

total cholesterol and diabetes (Table 3, model 3). The

equivalent risk estimates for the sample that included

those with pre-existing CVD was similar, with a HR of

1.1 (95% CI 0.70 - 1.80) for microalbuminuria and 3.8

(95% CI 2.37 - 6.24) for macroalbuminuria. The risk was

lowest for people with macroalbuminuria in the first

three years. For example, those people with macroalbu-

minuria that survived the first three years had a crude HR

of an incident cardiovascular event or death of 13.0 (95%

CI 6.45 - 26.39) to a peak of 32.3 (95% CI 8.55 - 121.77)

for those who survived the first five years (multivariate

adjustment not undertaken due to small event numbers).

The hazard dropped in the 6th year although this estimate

is based on small numbers.

4. DISCUSSION

Macroalbuminuria increased the risk of an incident

cardiovascular event or death, with increasing magnitude

over time, ranging from a HR of 9.0 for those who sur-

vived the first year to HR 32.3 for those who survived

the first five years. These estimates were attenuated by

age, gender and the co-variates added to the model, but a

strong independent association with cardiovascular dis-

ease remained. The estimates that included those with

pre-existing CVD were similar in magnitude to those for

incident CVD, suggesting that albuminuria may be a

relevant clinical indicator in both primary health care and

specialist medical settings. Microalbuminuria also con-

sistently elevated the risk of a cardiovascular event or

death compared with normal (less so than macroalbu-

minuria), but did not appear to follow the same time

varying pattern as macroalbuminuria. These findings

suggest that the risks associated with macroalbuminuria

are less in the first three years than the overall hazard

would suggest and so suggest that a window of opportu-

nity exists to reduce incident cardiovascular disease and

a subsequent CVD event or related death for those with

pre-existing CVD before the risk increases up to three

fold over the next three years.

The hazard ratio for those who survived the first six

years was a third of that for the previous five year esti-

mate. This difference may reflect small numbers of

events in the seventh year and so be an unstable and

non-valid estimate. However, if this is a true finding it

suggests that macroalbuminuria is a strong predictor of

incident cardiovascular disease in the short term (within

the first six years) and less so in the long term. This

finding is consistent with a study conducted in 1993

K. D’Onise et al. / Open Journal of Preventive Medicine 3 (2013) 22-27

Table 3. Level of albuminuria on the risk of cardiovascular event or death, Well Person’s Health Check, 1999-2006.

Incident cardiovascular disease, n = 2349 Including pre-existing cardiovascular disease, n = 2445

Model 1: Crude

Hazard Ratio,

(95% CI‡)

Model 2†:

Hazard Ratio,

(95% CI)

Model 3: Hazard

Ratio, (95% CI)

Model 1: Crude

Hazard Ratio,

(95% CI)

Model 2: Hazard

Ratio, (95% CI)

Model 3:

Hazard Ratio,

(95% CI)

All Years Micro§ 3.0 (1.83 - 4.96) 1.6 (0.97 - 2.73)1.3 (0.78 - 2.31)2.6 (1.65 - 3.97)1.4 (0.86 - 2.13) 1.1 (0.70 - 1.80)

Macro§ 10.8 (6.58 - 17.68) 6.1 (3.66 - 10.15)4.4 (2.52 - 7.83)9.7 (6.38 - 14.82)5.3 (3.45 - 8.22) 3.8 (2.37 - 6.24)

1 Year¶ Micro 3.0 (1.76 - 5.28) 1.6 (0.9102.87) 1.3 (0.71 - 2.33)2.7 (1.66 - 4.42)1.4 (0.85 - 2.32) 1.13 (0.67 - 1.92)

Macro 9.0 (5.03 - 15.98) 5.1 (2.79 - 9.16)3.5 (1.82 - 6.75)8.1 (4.90 - 13.43)4.5 (2.66 - 7.48) 3.1 (1.72 - 5.41)

2 Years Micro 2.7 (1.45 - 4.97) 1.4 (0.74 - 2.67)1.2 (0.60 - 2.29)2.5 (1.43 - 4.44)1.3 (0.72 - 2.30) 1.1 (0.59 - 1.98)

Macro 9.5 (5.14 - 17.66) 5.4 (2.84 - 10.09)4.1 (2.05 - 8.29)9.7 (5.64 - 16.78)5.3 (3.05 - 9.32) 4.0 (2.14 - 7.39)

3 Years Micro 2.7 (1.27 - 5.82) 1.6 (0.71 - 3.41)1.4 (0.63 - 3.20)2.5 (1.23 - 4.91)1.4 (0.68 - 2.82) 1.2 (0.58 -2.55)

Macro 13.0 (6.45 - 26.39) 7.9 (3.84 - 16.36)6.9 (3.07 - 15.43)10.2 (5.32 - 19.72)6.2 (3.17 - 12.08) 4.9 (2.33 - 10.51)

4 Years Micro 3.8 (1.38 - 10.52) 2.2 (0.77 - 6.30)2.2 (0.75 - 6.46)3.0 (1.20 - 7.27)1.3 (0.57 - 3.17) 1.6 (0.62 - 4.17)

Macro 21.1 (8.31 - 53.65) 13.1 (5.02 - 34.01)11.8 (4.14 - 33.59)14.9 (6.55 - 34.10)7.2 (3.38 - 15.23) 7.7 (3.01 - 19.86)

5 Years Micro 3.1 (0.63 - 15.56) 1.7 (0.34 - 8.95)- 1.9 (0.44 - 7.80)1.0 (0.23 - 4.32) -

Macro 32.3 (8.55 - 121.77) 19.2 (4.94 - 74.41)- 20.3 (6.81 - 60.76)11.8 (3.88 - 36.23) -

6 Years Micro 2.6 (0.16 - 42.00) - - 1.3 (0.12 - 14.40)- -

Macro 10.1 (0.63 - 161.1) - - 4.80 (0.43 - 52.91)- -

†Model 2: adjusted for age and gender; Model 3: also adjusted for baseline smoking, systolic blood pressure, total cholesterol and diabetes. ‡CI confidence

interval. §Micro—microalbuminuria, Macro—macroalbuminuria. ¶Each year denotes the number of events excluded—e.g. 1 year analyses all data excluding

those who had a cardiovascular event or death in the previous 1 year, 2 year analyses all data excluding those who had a cardiovascular event or death in the

first 2 years, etc.

where after 5 years the predictive effect of albuminuria

on the risk of CVD had reduced to no effect [12]. This

suggests the possibility of different trajectories of risk for

different potential causal pathways between macroalbu-

minuria and cardiovascular disease, with one or more

trajectories leading to a short term high increased risk of

death and a possible other whereby macroalbuminuria is

less predictive of cardiovascular disease from the start.

Another possible explanation is a healthy survivor effect,

where those who survive or do not develop cardiovascu-

lar disease by five years being healthier (in ways for

which we were unable to adjust) and so there is an ap-

parent reduction in the associated risk with cardiovascu-

lar disease by the seventh year. This warrants further

exploration in other cohorts as it may help to understand

the pathological processes involved by examining defin-

ing characteristics of the different cohorts (survivors

versus those who developed cardiovascular disease).

The findings from this study may not be generalisable

to other populations given the high prevalence of albu-

minuria and other CVD risk conditions especially diabe-

tes in this cohort relative to the general Australian popu-

lation. The magnitude of the risk estimates was also

higher than has been commonly reported in the literature

[2,4], although the reduction in risk for in the first few

years relative to the overall hazard may be generalisable.

There was also no difference in cardiovascular disease

risk for women compared with men, a finding that has

been reported from this cohort previously. This is

thought to be a result of higher rates of metabolic disease

especially diabetes in women compared with men and so

eliminating the protective effect of being female [7].

The findings of this study are possibly limited by un-

der-enumeration of outcome events. This could have

occurred through the data linkage process as it was un-

dertaken manually, and it may also be that deaths were

under-recorded in the latter years of the study given the

time-lag involved in registering deaths and cause of

death. These possible misclassification errors are likely

to have been non-differential and so are likely to have

had an effect of biasing the findings towards the null. For

example, if an assumption is made that the sensitivity of

the classification method was 95% and the specificity

99% (an ideal test), then the biasing effect on the esti-

mate would be an increase in the relative risk from 9.7

(the equivalent of the HR of 10.8) to 19.0 [13]. The se-

lection method for the first half of the cohort lead to

fewer people being tested for ACR than the latter half of

the cohort and so the results reflect that for a relatively

high risk cohort. While this introduces a potential risk of

K. D’Onise et al. / Open Journal of Preventive Medicine 3 (2013) 22-27 27

selection bias, there was no evidence for this on

re-analysis of the data restricted to only those communi-

ties where ACR testing was universal (16 of the 26

communities). There was minimal change in the effect

estimates, although they were measured with less preci-

sion (data not presented, available on request from the

authors).

In summary, the findings here suggest that the risk of

CVD with macroalbuminuria are not linear over time,

and that the first three years after finding macroalbumin-

uria provide a window of opportunity to manage the risk

of both incident and pre-existing cardiovascular disease

before the risk of a CVD related event or death elevates

substantially in subsequent years. These findings to-

gether suggest the need for increased use of ACR testing

in the course of clinical care for adults, including those

who are apparently at low risk of CVD.

This might represent an opportunity to step up primary

health care-based clinical intervention, especially aggres-

sive management of albuminuria with a view to revers-

ing or reducing subsequent CVD and other events. Fur-

ther, clinicians identifying an elevated ACR should con-

sider a broader assessment of the patient and manage-

ment of potential cardiovascular risk. This change in prac-

tice could be monitored in cohorts over time utilising

new capabilities in unit level record linkage.

5. ACKNOWLEDGEMENTS

Thanks to the participants of the WPHC study, Rohan Pratt for data

extraction, and staff from the Torres Strait and Northern Peninsula Area

Health Service District for ongoing collaboration. This study was

funded by National Health and Medical Research Council project

grants.

REFERENCES

[1] Basi, S., Fesler, P., Mimran, A. and Lewis, J.B. (2008)

Microalbuminuria in type 2 diabetes and hypertension.

Diabetes Care, 31 , S194-S201. doi:10.2337/dc08-s249

[2] Chronic Kidney Disease Prognosis Consortium (2010)

Association of estimated glomerular filtration rate and

albuminuria with all-cause and cardiovascular mortality

in general population cohorts: A collaborative meta-

analysis. Lancet, 375, 2073-2081.

doi:10.1016/S0140-6736(10)60674-5

[3] van der Velde, M., Matsushita, K., Coresh, J., Astor, B.C.,

Woodward, M., Levey, A. de Jong, P., Gansevoort, R.T.

and Chronic Kidney Disease Prognosis Consortium (2011)

Lower estimated glomerular filtration rate and higher al-

buminuria are associated with all-cause and cardiovascu-

lar mortality. A collaborative meta-analysis of high-risk

population cohorts. Kidney International, 79, 1341-1352.

doi:10.1038/ki.2010.536

[4] Pedrinelli, R., Dell’Omo, G., Penno, G. and Mariani, M.

(2001) Non-diabetic microalbuminuria, endothelial dys-

function and cardiovascular disease. Vascular Medicine, 6,

257-264. doi:10.1177/1358836X0100600410

[5] Ussai, K., Keith, S., Pequignot, E. and Falkner, B. (2011)

Risk factors associated with urinary albumin excretion in

African Americans. Journal of Human Hypertension, 25,

3-10. doi:10.1038/jhh.2010.79

[6] Miller, G., McDermott, R., McCulloch, B., Leonard, D.,

Arabena, K. and Muller, R. (2002) The well persons

health check: A population screening program in indige-

nous communities in North Queensland. Australian

Health Review, 25, 140-151. doi:10.1071/AH020136b

[7] McDermott, R., McCulloch, B. and Li, M. (2011) Gly-

caemia and albuminuria predict excess incident coronary

heart disease in Aboriginal and Torre Strait Islander

adults: A North Queensland cohort. Medical Journal of

Australia, 194, 514-518.

[8] Munoz, A. and Gange, S.J. (1998) Methodological issues

for biomarkers and intermediate outcomes in cohort stud-

ies. Epidemiology Reviews, 20, 29-42.

doi:10.1093/oxfordjournals.epirev.a017970

[9] Pinto-Sietsma, S.J., Mulder, J., Janssen, W.M.T., Hillege,

H.L., de Zeeuw, D. and de Jong, P.E. (2000) Smoking is

related to albuminuria and abnormal renal function in

nondiabetic persons. Annals of Internal Medicine, 133,

585-591.

[10] Cirillo, M., Senigalliesi, M., Laurenzi, M., Alfieri, R.,

Stamler, J., Stamler, R., Panarelli, W. and De Santo, N.G.

(1998) Microalbuminuria in nondiabetic adults: Relation

of blood pressure, body mass index, plasma cholesterol

levels, and smoking: The Gubbio population study. Ar-

chives of Internal Medicine, 158, 1933-1939.

doi:10.1001/archinte.158.17.1933

[11] StataCorp (2007) Stata statistical software: Release 10.

StataCorp LP, College Station.

[12] Damsgaard, E., Frøland, A., Jørgensen, O. and Mogensen,

C. (1993) Prognostic value of urinary albumin excretion

rate and other risk factors in elderly diabetic patients and

non-diabetic control subjects surviving the first 5 years

after assessment. Diabetologia, 36, 1030-1036.

doi:10.1007/BF02374495

[13] Lash, T., Fox, M. and Fink, A. (2009) Applying quantita-

tive bias analysis to epidemiologic data. Springer, New

York. doi:10.1007/978-0-387-87959-8