clinical investigation

http://www.kidney-international.org
& 2014 International Society of Nephrology

see commentary on page 682

Modifiable lifestyle and social factors affect chronic

kidney disease in high-risk individuals with type 2
diabetes mellitus
Daniela Dunkler1,2,3, Maria Kohl1,2,3, Georg Heinze3, Koon K. Teo1, Annika Rosengren4, Janice Pogue1,
Peggy Gao1, Hertzel Gerstein1, Salim Yusuf1, Rainer Oberbauer3,5 and Johannes F.E. Mann1,2,6 for the
ONTARGET Investigators
1

Population Health Research Institute, Hamilton Health Sciences/McMaster University, Hamilton, ON, Canada; 2University of ErlangenNurnberg, Department of Nephrology, Erlangen, Germany; 3Section for Clinical Biometrics, Center for Medical Statistics, Informatics and
Intelligent Systems, Medical University of Vienna, Vienna, Austria; 4Sahlgrenska University Hospital/Ostra Hospital, Goteborg, Sweden;
5
KH Elisabethinen Linz and Department of Internal Medicine III, Medical University of Vienna, Vienna, Austria and 6Schwabing General
Hospital, and KfH Kidney Center, Munich, Germany

This observational study examined the association between

modifiable lifestyle and social factors on the incidence and
progression of early chronic kidney disease (CKD) among
those with type 2 diabetes. All 6972 people from the Ongoing
Telmisartan Alone and in Combination with Ramipril Global
Endpoint Trial (ONTARGET) with diabetes but without
macroalbuminuria were studied. CKD progression was
defined as decline in GFR of more than 5% per year,
progression to end-stage renal disease, microalbuminuria, or
macroalbuminuria at 5.5 years. Lifestyle/social factors
included tobacco and alcohol use, physical activity, stress,
financial worries, the size of the social network and education.
Adjustments were made for known risks such as age, diabetes
duration, GFR, albuminuria, gender, body mass index, blood
pressure, fasting plasma glucose, and angiotensin-converting
enzyme inhibitors/angiotensin-receptor blockers use.
Competing risk of death was considered. At study end, 31%
developed CKD and 15% had died. The social network score
(SNS) was a significant independent risk factor of CKD and
death, reducing the risk by 11 and 22% when comparing the
third to the first tertile of the SNS (odds ratios of CKD 0.89 and
death 0.78). Education showed a significant association with
CKD but stress and financial worries did not. Those with
moderate alcohol consumption had a significantly decreased
CKD risk compared with nonusers. Regular physical activity
significantly decreased the risk of CKD. Thus, lifestyle is a
determinant of kidney health in people at high cardiovascular
risk with diabetes.
Kidney International (2015) 87, 784791; doi:10.1038/ki.2014.370;
published online 10 December 2014

Correspondence: Johannes F.E. Mann, Population Health Research Institute,

Hamilton Health Sciences/McMaster University, 426, Barton Street East,
Hamilton, ON, Canada. E-mail: prof.j.mann@googlemail.com
Received 9 June 2014; revised 9 September 2014; accepted 11
September 2014; published online 10 December 2014
784

The prevalence of diabetes mellitus increased rapidly over

the past decades to about 10% in industrialized, and even
more so in developing, countries to become a global health
problem.1,2 Roughly one-third of individuals with diabetes
develop chronic kidney disease (CKD) and up to 50% of
the dialysis population is diabetic.3 Drugs do not prevent
but rather mitigate diabetic CKD, they are costly, and they
exhibit adverse effects. It is therefore sensible to search for
modifiable factors, including lifestyle, associated with
progression of diabetic CKD, especially in early stages. In
fact, some lifestyle factors contribute to the development
of diabetes itself and to its huge burden of cardiovascular
disease.4
Although there is considerable literature on the impact of
lifestyle/social factors on cardiovascular outcomes and on
mortality, there is not much evidence of lifestyle/social factors
and early diabetic kidney disease. Some evidence associates
more advanced kidney disease with unhealthy behaviors,
lower social class, and poverty, and we recently reported that
individuals with diabetes eating a healthy diet and drinking
alcohol moderately develop less kidney damage.58 There is
also modest data relating progression of advanced CKD with
alcohol intake and physical activity.9
Diet, however, is only one aspect of lifestyle, a term first
coined in 1929 by Alfred Adler, an Austrian physician and
psychotherapist.10 Its quantification is not straightforward,
and it requires the measurement of a whole array of parameters from different domains such as nutrition, smoking,
bodily appearance, physical activity, stress, worries, work,
wealth, education, mobility, and social contacts/networks.
In the Ongoing Telmisartan Alone and in Combination
with Ramipril Global Endpoint Trial (ONTARGET), many of
these aspects of lifestyle were documented at baseline, as were
parameters of CKD, during the course of the 5.5-year trial.
Therefore, we thought to determine, without prespecified
hypotheses, which lifestyle/social factors associate with
Kidney International (2015) 87, 784791

D Dunkler et al.: Lifestyle/social factors and kidney disease in diabetes

clinical investigation

incidence or progression of early diabetic CKD to unravel

clues for future intervention studies.

Figure 2). Risk of CKD was 11% lower in the third tertile
of SNS compared with the first tertile (ORCKD3vs1 0.89
(95% confidence interval, 0.810.97)). Mortality was similarly
reduced with higher SNSs (Po0.001; ORdeath3vs1 0.78 (CI,
0.690.88)). Adjustment for alcohol intake, tobacco use,
physical activity, educational level, and Mini-Mental State
Examination score (MMSE), in addition to the above confounders, did not alter the association between SNS and
CKD. No significant pairwise interaction between SNS and
gender was found (P 0.325). The individual subcomponents of the SNS had similar associations with CKD and
mortality, although not always statistically significant.
Education showed an association with CKD (P 0.036).
Compared with participants without education, participants
with higher education showed a graded decrease in the risk of
CKD (college/university versus no education ORCKD 0.65
(CI, 0.480.88)).
Alcohol was regularly consumed by 32% (n 2250) of
participants, 30% (n 2070) consumed quantities within
WHO criteria of moderate drinking, and 2% (n 180)
exceeded these. Those drinking moderately had a significantly
reduced risk of CKD compared with nonusers in the model
adjusted for the predefined confounders listed in Statistical
Analysis (P 0.001; ORCKD 0.79 (CI, 0.690.89); Table 2).
Physical activity ranged from mainly sedentary (n 1677,
24%) to being physically active every day (n 2406, 35%). In
adjusted models, more regular physical activity significantly
reduced the risk of CKD (P 0.009), as well as mortality
(Po0.001). Participants who were physically active every day
had a significantly reduced risk of CKD compared with participants being mainly sedentary (ORCKD 0.76 (CI, 0.650.88)).
Tobacco was never used by 40% (n 2773) of participants, 49% (n 3426) had stopped using tobacco before
inclusion to the study, and 11% (n 764) were still smoking.
There was no association with CKD (P 0.943), but a
significant increase in mortality (Po0.001). Therefore, an
explanation of the first might be the competing risk of death.
In the whole ONTARGET population with diabetic and
nondiabetic participants and a very similar distribution of
the use of tobacco, a significant increase in CKD (P 0.005)
was detected (current versus never users ORCKD 1.19 (CI,
1.06-1-35)) (Supplementary Table S4 and Figure S2 online).
Two factors did not exhibit an association with the composite renal outcome: stress (home or work) did not alter
the risk of CKD (P 0.302), as well as financial worries
(P 0.451).

RESULTS

Of 6972 participants with type 2 diabetes mellitus and

normoalbuminuria or microalbuminuria at baseline, 2182
(31%) experienced incidence or progression of CKD, the
primary composite renal end point, and 1028 (15%) had died
(Figure 1). Namely, 1477 (21%) participants experienced a
decline in estimated glomerular filtration rate (GFR) of more
than 5% per year, 974 (14%) participants developed new
microalbuminuria or macroalbuminuria, 39 (0.6%) progressed
to end-stage renal disease, and 269 (4%) experienced at least
two renal end points (Supplementary Table S1 online).
Participants characteristics and lifestyle data are depicted in
Table 1.
The following lifestyle/social factors showed a significant
association with the composite renal end point: The physical
social network, defined as the number of social interactions
and personal relationships (social network score, SNS),
education, alcohol intake, and physical activity (Supplementary Table S2 and Figure S1 online).
The median SNS was 16 people with an interquartile range
from 9 to 28. Participants consuming alcohol, with higher
body mass index (BMI), without financial worries, with
better education, being physically active at least once a week,
male participants, and former and current tobacco users
showed on average higher SNSs (Supplementary Table S3
online). The adjusted model showed a significantly decreased
risk of CKD for participants with higher SNS (Po0.001;

ONTARGET n = 25,620

18,648 participants were excluded:

16001 without diabetes,
823 with missing UACR at baseline,
758 with macroalbuminuria at baseline,
17 with missing social network score at baseline, and
1049 with missing outcome.

Participants at baseline n = 6972 (100%)

Primary renal outcome

n = 2182 (31%)

Death
n = 1028 (15%)

GFR-decline outcome: n = 1477 (21%)

Albuminuria outcome: n = 974 (14%)
ESRD: n = 39 (0.6%)

Figure 1 | Flow chart of the number of participants for the

primary outcome after 5.5 years of follow-up. GFR at baseline was
obtained for all participants. Of the 1049 participants with missing
outcomes, 387 had no UACR measurement, 20 had no GFR
measurement, and 642 had neither measurement at the last
follow-up visit of the study, but they were alive. ESRD, end-stage
renal disease; GFR, glomerular filtration rate; UACR, urinary
albumincreatinine ratio.
Kidney International (2015) 87, 784791

Sensitivity analyses

Analysis of the whole ONTARGET population (n 19,756)

with 4704 (24%) composite renal outcomes essentially confirmed the results from the diabetic subset (Supplementary
Table S3 and Figure S2 online). The risk of CKD was
significantly reduced for participants consuming moderate
quantities of alcohol (P40.001; ORCKD 0.83 (CI, 0.770.90)),
for participants being physically more active (P 0.001),
or for participants with a higher SNS (Po0.001; ORCKD3vs1
785

clinical investigation

D Dunkler et al.: Lifestyle/social factors and kidney disease in diabetes

Table 1 | Clinical and lifestyle characteristics of participants

with type 2 diabetes mellitus

Table 1 (Continued )

Characteristics at baseline

No.a

Median (IQR), mean

(s.d.) or no. (%)

Ethnic groupCaucasian
Genderfemale
Age
BMI
Waist circumference (cm)

6972
6972
6972
6939
6958

4738 (67.96)
2229 (31.97)
66 (6171)
29.22 (4.89)
99.01 (13.56)

Blood pressure (mmHg)

Systolic
Diastolic
Mean arterial
Diabetes duration (years)
Previous ACEI/ARBs
Glucose (mmol/l)
UACR (mg/mmol)
CKD-EPI GFR ml/min/1.73 m2
Serum creatinine (mmol/l)
Microalbuminuria

6968
6968
6968
6912
6972
6939
6972
6956
6972
6972

142.99 (16.73)
81.68 (10.22)
102.12 (11.04)
8.5 (316)
5000 (71.72)
7.78 (6.409.86)
6.58 (2.8525.11)
71.25 (18.09)
88.40 (76.91106.08)
1589 (22.79)

6972

ONTARGET randomization
Ramipril
Telmisartan
Combination
Mini-mental state examination

6898

2382 (34.17)
2270 (32.56)
2320 (33.28)
28 (2630)

Clinical history
Cardiovascular diseaseb
Myocardial infarction
Peripheral artery disease
Stroke or TIA
Coronary artery disease
Hypertension
CABG
PTCA

6972
6972
6972
6972
6972
6972
6972
6972

4128 (59.21)
2702 (38.76)
999 (14.33)
1267 (18.17)
4412 (63.28)
5474 (78.51)
1358 (19.48)
1576 (22.60)

Medications
Statin
b-Blocker
Any antiplatelet drug
Diuretic
Calcium-channel blocker

6972
6972
6972
6972
6972

3847
3574
5216
2447
2547

Lifestyle & social factors

Social network score

6972

16 (927.50)

6945
6967
6944
6957
6970

4 (210)
3 (26)
3 (16)
3 (26)
4720 (67.72)
2070 (29.70)
180 (2.58)
2773 (39.82)
3426 (49.20)
764 (10.97)
1677 (24.06)
815 (11.69)
1595 (22.88)
477 (6.84)
2406 (34.52)
3016 (43.28)
3062 (43.94)
891 (12.79)
4615 (66.20)

Number of peopley
You know and interact withc
You can speak frankly withd
Who can drop bye
You visit or visit youf
Alcohol intakenone
Within WHO criteria
Above WHO criteria
Tobacco usenever
Former
Current
Physical activitymainly sedentary
p1 per week
24 per week
56 per week
Everyday
Stress at home or worknever
Some
Severe/permanent
Worry about moneynone/little

786

6963

6970

6969

6971

(55.18)
(51.26)
(74.81)
(35.10)
(36.53)

Characteristics at baseline
Moderate
High/severe
Formal educationnone
18 years
912 years
Trade/technical school
College/university

No.a

6971

Median (IQR), mean

(s.d.) or no. (%)
1836
520
269
2256
2006
1146
1294

(26.34)
(7.46)
(3.86)
(32.36)
(28.78)
(16.44)
(18.56)

Abbreviations: ACEI/ARBs, angiotensin-converting enzyme inhibitors/angiotensinreceptor blockers; BMI, body mass index; CABG, coronary artery bypass graft; CKDEPI, Chronic Kidney Disease Epidemiology Collaboration equation; GFR, glomerular
filtration rate; IQR, interquartile range; ONTARGET, Ongoing Telmisartan Alone and
in Combination with Ramipril Global Endpoint Trial; PTCA, percutaneous transluminal coronary angioplasty; TIA, transient ischemic attack; UACR, urinary albumin
creatinine ratio.
SI conversion factors: to convert creatinine to mg/dl, multiply by 1/88.4; to convert
glucose to mg/dL, multiply by 18.15.
Median (interquartile range, IQR), mean (standard deviation, s.d.), or frequencies
(percentages) are given depending on the type and distribution of the respective
characteristics.
a
Number of participants with available data.
b
Cardiovascular disease was defined as history of myocardial infarction, peripheral
artery disease, and/or stroke/TIA.
c
About how many people do you know personally and interact with who share
your interests?
d
About how many friends or family members do you have with whom you can
speak frankly?
e
About how many friends do you have who would drop by your home
unexpectedly and you wouldnt be embarrassed if your home was untidy?
f
Not including those who live with you, about how many people do you visit or
visit you in an ordinary week?

0.88 (CI, 0.830.93)). Higher education showed a graded

decrease in the risk of CKD (P 0.001) and mortality
(Po0.001). Financial worries were significantly associated
with CKD (P 0.005; moderate versus no worries ORCKD
1.14 (CI, 1.051.24)).
Models adjusted with the reduced set of confounders
(defined in Statistical Analysis) provided very similar
estimates (Supplementary Tables S5 and Figure S3 online).
The combined outcome of CKD and death (Supplementary
Table S6 and Figure S4 online), as well as separate analyses of
GFR decline (Supplementary Table S7 and Figure S5 online)
and albuminuria (Supplementary Table S8 and Figure S6
online), showed similar associations as for the primary
composite renal end point. Accordingly, the omission of
potentially very sick individuals, i.e., being mainly sedentary
or with a MMSE X 24, did not change the results (Supplementary Table S910 and Figure S78 online).
DISCUSSION

By analyzing a range of lifestyle and social factors, we observed

an association of several such factors with the incidence and
progression of early diabetic CKD, including the size of the
social network (more social relations preventive), and education (higher education preventive), alcohol intake (moderate
intake preventive), and physical activity (higher activity
preventive). Mortality was examined as competing risk, and,
in general, similar associations were found as for CKD.
Kidney International (2015) 87, 784791

clinical investigation

D Dunkler et al.: Lifestyle/social factors and kidney disease in diabetes

Death

300
200
100
0

1.6

30.78%

1.4
1.2
1.0
0.8

600
400
200
0
0

20

40
60
Social network score

80

16.82% 14.66%

1.6

12.74%

1.4
1.2
1.0
0.8
0.6
Frequency

0.6

Relative odds of death

32.02% 31.46%

Frequency

Relative odds of
renal endpoint

Incidence or progression of CKD

20

40
60
Social network score

80

Figure 2 | Plots of relative odds for the social network score (SNS) adjusted with the set of confounders. Association of the SNS and
relative odds (solid line) with 95% confidence interval (dashed lines) with incidence or progression of early diabetic chronic kidney disease (left),
or death (right). Histograms show the distribution of the SNS in the respective outcome state. The SNS ranges from 0 to 100%, with 1%
approximately equaling 1.1 people one regularly interacts with. Gray vertical lines display tertiles, and the number within each tertile gives the
percentage of participants experiencing the respective outcome state. For further details, see legend to Table 2.

The absence of a social network independently increased

the risk of CKD, a new and surprising finding. Although the
association with CKD has, to our knowledge, not been
reported, impressive evidence links having fewer social
relations with high mortality, as recently reviewed by Tay
et al.11 In other reports, dimensions of SNS such as low social
integration and low emotional support were highly predictive
for incident coronary disease and mortality.12 One may consider
that a tight social network helps in the (self-)management of
a chronic disease such as diabetes and therefore prevents
complications.13 Intriguingly, there may also be a biological
explanation, as low social support was associated with shorter
telomere length in older persons with diabetes.14 Thus, social
environment may contribute to cellular aging particularly in
later life. Despite the established associations of SNS and
cardiovascular events and of the latter with CKD,15 the
association of SNS with CKD was not examined yet, certainly
not for early diabetic CKD.
Previous longitudinal studies investigated parameters of
socioeconomic status and renal outcomes, and several studies
reported low socioeconomic status or poverty as a risk for
low GFR, albuminuria, and even end-stage renal disease, but
none of the studies separated individual characteristics
from the area socioeconomic background or focused on
early diabetic CKD.5,16,17 In line is the recent finding from
the REGARDS trial that household, but not community,
poverty is independently associated with CKD.7 Interestingly,
investigators of the Atherosclerosis Risk in Communities
study showed that living life-long in a disadvantaged neighborhood of the United States was not associated with a higher
risk of developing a GFR below 45 ml/min.18 Nevertheless,
there was an association of CKD with working class status.
A German cross-sectional study of individuals with diabetes
and CKD described no substantial association of socioeconomic status with prevalent GFR or albuminuria, but
important further risks for CKD such as HbA1c, diabetes
duration, and BMI were not analyzed.19
Moderate alcohol intake and higher education were singled
out as lifestyle/social factors with substantial association with
Kidney International (2015) 87, 784791

early diabetic CKD. There is only modest and contradictory

evidence about alcohol consumption and decline in renal
function.9 The present data may explain some of the contradictions, as we found only moderate, but not heavy, drinking
to be beneficial for renal outcomes. One can speculate that
moderate amounts of alcohol are protective for the renal
vasculature, whereas higher intake promotes high blood
pressure.9 However, moderate alcohol consumption may also
be an indicator of social integration and overall well-being.
This hypothesis is supported by our finding that the highest
tertile of the SNS included the highest number of individuals
with moderate alcohol intake.
Traditionally, lifestyle intervention programs tackle diet
and physical activity. Unfortunately, long-term results are
disappointing even in participants of well-conducted intervention studies.20 Thus, obesity and sedentary behavior,
classical lifestyle factors, were not sustainably modifiable in
several trials,20,21 and they did not alter the incidence of
advanced diabetic CKD.22 Our results, however, suggested
that engaging in social networks may be an alternative intervention, if the identified associations would be causal. For
example, comparing 1000 diabetic individuals with 25 contacts
with those with only 5 contacts, an additional 14 individuals
would be protected from CKD in the former group. We also
found a clear association of physical social networks with
mortality, extending previous studies in general population
to diabetes. In a previous analysis, we showed that individuals with diabetes adhering to a healthier diet measured by
the modified Alternate Healthy Eating Index (mAHEI) had a
reduced risk of CKD and mortality.6 There is no association
between the mAHEI and the SNS (Pearson correlation 0.14).
In accordance with a recent paper by Robinson-Cohen
et al. who analyzed the 256 participants of the Seattle Kidney
Study, we found that physical activity was inversely associated
with CKD incidence and progression.23 Our finding was
independent of BMI, which is a factor that has been shown to
cause glomerular hypertension and to stimulate renal fibrosis.
In individuals with diabetes, education showed a significant
association with a reduced risk of CKD, but it had no impact
787

Never
Never
None/little

None
Mainly sedentary

Tobacco use
Stress at home or work
Financial worries

Alcohol intake
Physical activity

Abbreviation: ACEI/ARB, angiotensin-converting enzyme inhibitor/angiotensin-receptor blocker; CKD, chronic kidney disease; GFR, glomerular filteration rate; ONTARGET, Ongoing Telmisartan Alone and in Combination with
Ramipril Global Endpoint Trial; UACR, urinary albumincreatinine ratio.
ORCKD and ORdeath compare two individuals with different characteristics in their odds to develop CKD or to die, respectively. ORs are given for the median of the 2nd and 3rd tertile, i.e., 50.0th and 83.3rd percentiles, compared
with the median of the 1st tertile, i.e., 16.7th percentile. P-values for incidence or progression of CKD and for death are stated for each variable. Independent variables in bold type have a significant association with incidence or
progression of early diabetic CKD.
The primary composite renal end point of incidence or progression of CKD after up to 5.5 years of follow-up was defined as occurrence of either a clinically relevant GFR decline or new microalbuminuria or macroalbuminuria.
The competing risk of death was considered.
The models are adjusted for the following confounders measured at baseline: age, duration of diabetes, GFR, status of albuminuria, gender, ONTARGET randomization arms, D-UACR to progression, which was defined as the
difference between the participant-specific cutoff point of developing a new microalbuminuria or macroalbuminuria and UACR at baseline on the log-scale, body mass index, mean arterial blood pressure, fasting plasma glucose,
and previous use of ACEI/ARBs.

o0.001
0.385
0.066
0.943
0.302
0.451

0.108
o0.001
0.001
0.009

0.036
o0.001
0.001
0.058
0.405
0.123
0.056
0.129
0.084
0.036

8
11
8
9
912 years
College/university
Above WHO criteria
24 per week
Everyday
Current
Severe/permanent
High/severe
3
4
3
3
18 years
Trade/technical school
Within WHO criteria
o1 per week
56 per week
Former
Some
Moderate
1
1
1
0
None

(0.7930.992)
(0.6820.865)
(0.6950.904)
(0.7051.006)
(0.5421.172)
(0.4801.068)
(0.4911.306)
(0.5540.845)
(0.4870.716)
(1.5862.600)
(0.7971.286)
(0.9961.735)
0.887
0.768
0.792
0.842
0.797
0.716
0.801
0.684
0.590
2.031
1.013
1.315
(0.9360.998)
(0.8490.940)
(0.8280.949)
(0.8301.003)
(0.5481.169)
(0.4691.062)
(0.7090.996)
(0.5990.994)
(0.4960.939)
(1.0321.467)
(0.7721.057)
(0.9731.369)
(0.9601.005)
(0.9311.001)
(0.9091.012)
(0.8751.008)
(0.5580.999)
(0.5240.974)
(0.6900.893)
(0.6851.015)
(0.6721.081)
(0.8971.160)
(0.8531.081)
(0.9551.231)

(0.8661.017)
(0.8461.002)
(0.8331.024)
(0.7791.016)
(0.5881.061)
(0.4770.881)
(0.6011.211)
(0.7130.984)
(0.6540.880)
(0.8431.254)
(0.7261.039)
(0.8091.242)
0.938
0.921
0.923
0.890
0.790
0.648
0.853
0.837
0.759
1.028
0.868
1.003
0.982
0.965
0.959
0.939
0.747
0.714
0.785
0.834
0.852
1.020
0.960
1.084

0.966
0.893
0.886
0.912
0.801
0.705
0.840
0.772
0.683
1.230
0.903
1.154

Death

o0.001
o0.001
34.70
16
7
0.779 (0.6920.876)
0.879 (0.8270.934)
0.885 (0.8070.972)

Median of tertiles/categories

2
1 (Reference)
ORdeath3vs1
ORdeath2vs1
ORCKD3vs1
ORCKD2vs1

D Dunkler et al.: Lifestyle/social factors and kidney disease in diabetes

0.939 (0.8950.985)

Social network score

Number of people
You visit or visit you
You know and interact
You can speak frankly with
Who can drop by
Formal education

P-Values

788

Continuous explanatory variables

Table 2 | Single-variable models adjusted with the set of confounders for the primary outcome at 5.5 years of follow-up

CKD

clinical investigation

on mortality. A likely interpretation of these findings is that

education helps cope with a chronic disorder such as diabetes,
but it has no effect on five-year mortality in patients with
advanced cardiovascular disease. This corresponds to the
findings of the comparable ADVANCE study.24 The investigators found that education, crudely measured as completion
of formal education before or after 16 years of age, was
associated with a reduced risk of CKD-related events.
For this analysis of 6972 participants with diabetes, we
accumulated over 30,000 patient-years with 2182 primary
renal outcomes and 1028 deaths. It is noteworthy that the
participants were treated to a high degree according to
guidelines with preventive cardiovascular drugs, that most of
them were at a target blood pressure, and that all received
angiotensin-converting
enzyme
inhibitors/angiotensinreceptor blockers. With this fairly large database, confidence
intervals were relatively narrow. Nevertheless, some limitations apply. This is a post hoc observational analysis of a large
randomized controlled trial, and it can only be hypothesisgenerating. To analyze outcomes, we included only participants with measurements of renal parameters at the beginning
and end of the study. Therefore, our cohort may differ from
the general ONTARGET population, namely those participants providing blood and urine samples at the study end
may have been more compliant. Kidney function was measured only at two times during follow-up, and therefore a
discrete time model was chosen for the renal end point after
5.5 years of follow-up. As there is complete follow-up, a timeto-event analysis, namely a Fine-Gray model, would not
materially alter results. We were concerned about multiple
testing of, and interdependence of, lifestyle factors. To gauge
the impact of a single factor, statistical techniques to address
confounding issues were incorporated and several sensitivity
analyses were conducted supporting main findings. Our
results are strengthened by the independence from confounders and possible mediators including age, diabetes
duration, GFR, albuminuria, drug treatment, obesity, and
glucose control. Another limitation of our analysis is that our
conclusions apply to individuals with diabetes and high
cardiovascular risk. This population contributes approximately
one-third of incident end-stage renal disease patients.25
However, in a sensitivity analysis, we confirmed the findings in the whole ONTARGET populationthat is, a cohort
still at high risk and with vascular disease but predominantly
without diabetes. We did not conduct our primary analysis
on the whole ONTARGET population because the clinical
relevance of early renal changes, namely changes in low-grade
albuminuria, are not established in nondiabetic CKD.
These associational results suggest that lifestyle factors
apart from dietary habits6 may be important risk factors of
incidence and progression of CKD in type 2 diabetes and
potentially in people with vascular disease. It appears that
as a new findinga high SNS, physical activity, and alcohol
in moderation are associated with less decrease in GFR and
less increase in albuminuria. Those findings may guide future
intervention trials.
Kidney International (2015) 87, 784791

D Dunkler et al.: Lifestyle/social factors and kidney disease in diabetes

clinical investigation

MATERIALS AND METHODS

Study design and population
The ONTARGET trial randomized 25,620 participants aged 55 years
or older, with vascular disease or type 2 diabetes mellitus with endorgan damage, from 40 countries to telmisartan, ramipril, or both
combined.26 The composite primary outcome was cardiovascular
mortality, myocardial infarction, stroke, or hospitalization for heart
failure. Secondary outcomes included doubling of serum creatinine,
need for dialysis, and changes in urine albumin. Main findings
included no difference in outcomes between the randomized groups
but more adverse events with combination therapy.27,28
For the present analysis, all 6972 ONTARGET participants with
type 2 diabetes mellitus and normoalbuminuria or microalbuminuria
at baseline were included. We excluded those with macroalbuminuria without urine albumin or serum creatinine measurement at
baseline or at last follow-up (Figure 1).
Urinary albumincreatinine ratio (mg/mmol; UACR) was measured
centrally at baseline and after 2 and 5 years of follow-up;29 serum
creatinine was measured locally at the same times. Estimated GFR
was calculated with the Chronic Kidney Disease Epidemiology
Collaboration equation.30

compared with baseline. The 30%-increase rule was included to

avoid random categorical changes for participants near a cutoff
point, and it decreased the number of participants with new microalbuminuria or macroalbuminuria only by 10. The competing risk
of death was considered, leading to the following primary outcome
after up to 5.5 years of follow-up: (1) alive without incidence or
progression of CKD, (2) alive with incidence or progression of CKD,
or (3) death.

Assessment of lifestyle factors

The following lifestyle/social factors were recorded at baseline:
alcohol intake, tobacco use (never, former, current), waist circumference (cm), physical activity (mainly sedentary, engaging in
physical activity p1, 24, 5-6/week, every day), stress at home or
work,31 financial worries (none/little, moderate, high/severe), and
formal education (none, 18, 912 years, trade/technical school,
college/university). Alcohol intake was categorized by the number of
drinks per week (one drink equaling 1.5 ounces of hard liquor, or
one glass of beer/wine), stratified by World Health Organization
criteria.32 Moderate alcohol intake was defined as 112 and 118
drinks/week, and heavy alcohol intake was defined as 412 and 418
drinks/week for women and men, respectively. Participants physical
social network, defined as the number of social interactions and
personal relationships, was assessed at baseline by four questions
quantifying the number of people one regularly interacts with. These
questions were as follows: (1) About how many people do you know
personally and interact with who share your interests? (2) About how
many friends or family members do you have with whom you can
speak frankly? (3) About how many friends do you have who would
drop by your home unexpectedly, and you wouldnt be embarrassed if
your home was untidy? and (4) Not including those who live with you,
about how many people do you visit or visit you in an ordinary week?
To quantify the size of a participants social network, a summary
score of the four social variables was devised (Supplementary Table
S11 online).
Study outcomes
The primary composite outcome was a renal end point addressing
the incidence or progression of CKD after up to 5.5 years of followup. This primary renal outcome was defined as occurrence of either
a clinically relevant GFR decline or new microalbuminuria or
macroalbuminuria. The former included a GFR decline of more
than 5%/year, as quantified by GFR measurements at baseline,
2 years, and 5.5 years, acute or chronic dialysis, or end-stage renal
disease; i.e., a GFR o15 ml/min/1.73 m2. New microalbuminuria or
macroalbuminuria were defined as progression of UACR above 3.4
or 33.9 mg/mmol, respectively, and an increase of at least 30%
Kidney International (2015) 87, 784791

Statistical analysis
Median and interquartile range, or mean and s.d., were used to
describe continuous variables, whereas frequencies and percentages
were used for categorical variables. Diabetes duration and UACR
were log-transformed because of skewed distributions. Multinomial
logit regression was applied to quantify the association between
outcomes and lifestyle/social factors. Nonlinear associations were
modeled using fractional polynomials (P 0.157).33 For the SNS
and the four questions quantifying the social network, only fractional polynomials of degree 1 were allowed. Pairwise interactions
between lifestyle/social factors and confounders were tested separately controlling for a false discovery rate of 5%. No clinically
meaningful and statistically significant interaction was detected.
A multinomial logit model yields two odds ratios (OR): ORCKD
quantifies the association of a variable with the risk of developing
CKD, and ORdeath quantifies the association of a variable with
mortality. For continuous variables, two ORs comparing the median
of the 2nd and 3rd tertile to the median of the 1st tertile, OR2vs1
and OR3vs1, are stated, respectively. For each variable, two Waldtest P-values for CKD and mortality are given.
On the basis of expert opinion, a set of confounders measured at
baseline was defined before analysis, including diabetes duration,
age, gender, albuminuria status, ONTARGET randomization arm,
GFR, and D-UACR to progression, which was defined as the difference between the participant-specific cutoff for diagnosis of new
microalbuminuria or macroalbuminuria and UACR at baseline on
the log-scale6 (BMI, kg/m2), fasting plasma glucose (mmol/L),
mean arterial blood pressure (mmHg), and previous use of
angiotensin-converting enzyme inhibitors/angiotensin-receptor
blockers (Supplementary Table S12 online). For each lifestyle/social
factor, a univariate model and a model adjusted with the
confounder set were developed as our main analysis. As lifestyle/
social factors are related to each other, we included those variables
only individually into the multivariable confounder model. Median
absolute Spearman correlation among all confounders was 0.05,
with the maximum correlation between age and GFR (0.39), and
median absolute correlation between confounders and lifestyle
factors was 0.04, with the maximum correlation between gender and
tobacco use (0.41).
To test robustness of findings, several sensitivity analyses were
conducted: (1) analyzing the whole ONTARGET population,
excluding those with macroalbuminuria or without urine albumin
or serum creatinine measurement at baseline or at last follow-up
(n 19,756; Supplementary Figure S9 online); (2) using a reduced
set of confounders where BMI, fasting plasma glucose, mean arterial
blood pressure, and the use of angiotensin-converting enzyme
inhibitors/angiotensin-receptor blockerss were excluded, because
they may be mediators rather than confounders; (3) using the
combined end point of CKD and death as outcome; (4) analyzing
GFR decline and albuminuria separately; and (5) under the
assumption that very sick individuals, i.e., mainly sedentary
789

clinical investigation

D Dunkler et al.: Lifestyle/social factors and kidney disease in diabetes

individuals or individuals with low MMSE scores, might modify

associations between lifestyle/social factors and CKD, the main
analysis was repeated for participants who were physical active
(n 5308, 76%) and for participants with an MMSE X 24
(n 6252, 90%).
Data quality was excellent, with missing baseline values in less
than 5% of participants. Therefore, complete case analysis was
applied. A two-sided P-value o0.05 was considered statistically
significant. Given the explanatory nature of the analysis, P-values
were not adjusted for multiple testing, but confidence limits are
provided. R 2.12.2 was used for analysis.34

Figure S7. Sensitivity analysis 5: Adjusted models for the primary

outcome with participants being physically active at least once a
week: Plots of relative odds.
Figure S8. Sensitivity analysis 5: Adjusted models with participants
with a Mini-Mental State Examination score X24: Plots of relative
odds.
Figure S9. Flow chart of the whole ONTARGET population.
Supplementary material is linked to the online version of the paper at
http://www.nature.com/ki

REFERENCES
1.

DISCLOSURE

All the authors declared no competing interests.

2.

ACKNOWLEDGMENTS

Funding for the ONTARGET trial was provided by Boehringer

Ingelheim. Funding for this study was provided by European
Unions 7 framework program (SysKid, a Collaborative FP7 Research
Project to Fight Chronic Kidney Disease; grant agreement number
HEALTHF22009241544) and by Population Health Research
Institute (PHRI), McMaster University, Hamilton, Canada. The funding
agency and Boehringer Ingelheim had no role in the design and
conduct of the study; in the collection, analysis, and interpretation of
the data; or in the preparation, review, or approval of the manuscript.

3.

4.

5.

SUPPLEMENTARY MATERIAL
eAppendix. The ONTARGET investigators listed by name, committee
and/or country.
Table S1. Primary outcome after up to 5.5 years of follow-up.
Table S2. Univariate analysis: Univariate models for the primary
outcome.
Table S3. Baseline characteristics of participants with diabetes
separated by tertiles of the social network score (SNS).
Table S4. Sensitivity analysis 1: Analysis of whole ONTARGET
population (n 19,756).
Table S5. Sensitivity analysis 2: Models adjusted with the reduced set
of confounders for the primary outcome.
Table S6. Sensitivity analysis 3: Adjusted models for the combined
outcome of renal end points and death.
Table S7. Sensitivity analysis 4: Adjusted models for the GFR-decline
outcome.
Table S8. Sensitivity analysis 4: Adjusted models for the albuminuria
outcome.
Table S9. Sensitivity analysis 5: Adjusted models for the primary
outcome with participants being physical active at least once a week
(n 5308).
Table S10. Sensitivity analysis 5: Adjusted models for the primary
outcome with participants with a Mini-Mental State Examination
score X24 (n 6252).
Table S11. Derivation of the social network score.
Table S12. Main analysis: Confounder model for the primary
outcome.
Figure S1. Univariate analysis: Univariate models for the primary
outcome: Plots of relative odds.
Figure S2. Sensitivity analysis 1: Analysis of whole ONTARGET
population: Plots of relative odds.
Figure S3. Sensitivity analysis 2: Models adjusted with the reduced
set of confounders for the primary outcome: Plots of relative odds.
Figure S4. Sensitivity analysis 3: Adjusted models for the combined
outcome of renal end points and death: Plots of relative odds.
Figure S5. Sensitivity analysis 4: Adjusted models for the GFR-decline
outcome: Plots of relative odds.
Figure S6. Sensitivity analysis 4: Adjusted models for the albuminuria
outcome: Plots of relative odds.
790

6.

7.

8.

9.
10.
11.

12.

13.

14.

15.
16.

17.

18.

19.

Yano Y, Fujimoto S, Asahi K, Watanabe T. Prevalence of chronic kidney

disease in China. Lancet 2012; 380: 213214.
Centers for Disease Control and Prevention. National diabetes fact sheet:
National estimates and general information on diabetes and prediabetes in
the United States. US Department of Health and Human Services, Centers
for Disease Control and Prevention: Atlanta, GA, USA, 2011; http://
www.cdc.gov/diabetes/pubs/estimates11.htm. Accessed 8 September
2014.
US Renal Data System. USRDS 2012 annual data report: Atlas of chronic
kidney disease and end-stage renal disease in the United States. National
Institutes of Health, National Institute of Diabetes and Digestive and
Kidney Diseases. http://www.usrds.org/atlas12.aspx. Accessed 8
September 2014.
Shi L, Shu XO, Li H et al. Physical activity, smoking, and alcohol
consumption in association with incidence of type 2 diabetes among
middle-aged and elderly chinese men. PLoS ONE 2013; 8: e77919.
Martins D, Tareen N, Zadshir A et al. The association of poverty with the
prevalence of albuminuria: data from the third National Health and
Nutrition Examination Survey (NHANES III). Am J Kidney Dis 2006; 47:
965971.
Dunkler D, Dehghan M, Teo KK et al. Diet and kidney disease in high-risk
individuals with type 2 diabetes mellitus. JAMA Int Med 2013; 173:
16821692.
Mc Clellan WM, Newsome BB, McClure LA et al. Poverty and racial
disparities in kidney disease: the REGARDS study. Am J Nephrol 2010; 32:
3846.
Muntner P, Judd SE, Gao L et al. Cardiovascular risk factors in CKD
associate with both ESRD and mortality. J Am Soc Nephrol 2013; 24:
11591165.
De Francisco ALM, Fresnedo GF, Palomar R et al. The renal benefits of a
healthy lifestyle. Kidney Int 2005; 68: S2S6.
Adler A. Understanding Human Nature, 1st edn Greenburgh, NY, USA 1927
pp 18.
Tay L, Tan K, Diener E, Gonzalez E. Social relations, health behaviors, and
health outcomes: a survey and synthesis. Appl Psychol Health Well Being
2013; 5: 2878.
Rosengren A, Wilhelmsen L, Orth-Gomer K. Coronary disease in relation
to social support and social class in Swedish men: a 15 year follow-up in
the study of men born in 1933. Eur Heart J 2004; 25: 5663.
Reeves D, Blickem C, Vassilev I et al. The contribution of social networks
to the health and self-management of patients with long-term
conditions: a longitudinal study. PLoS One 2014; 9: e98340.
Carroll JE, Diez Roux AV, Fitzpatrick AL, Seeman T. Low social support is
associated with shorter leukocyte telomere length in late life: multiethnic study of atherosclerosis. Psychosom Med 2013; 75: 171177.
Schiffrin EL, Lipman ML, Mann JFE. Chronic kidney disease: effects on the
cardiovascular system. Circulation 2007; 116: 8597.
Krop JS, Coresh J, Chambless LE. A community-based study of
explanatory factors for the excess risk for early renal function decline in
blacks vs whites with diabetes: the atherosclerosis risk in communities
study. Arch Intern Med 1999; 159: 17771783.
Klag MJ, Whelton PK, Randall BL et al. End-stage renal disease in
African-American and white men: 16-year MRFIT findings. JAMA 1997;
277: 12931298.
Shoham DA, Vupputuri S, Kaufman JS et al. Kidney disease and the
cumulative burden of life course socioeconomic conditions: the
Atherosclerosis Risk in Communities (ARIC) Study. Soc Sci Med 2008; 67:
13111320.
Wolf G, Busch M, Muller N, Muller UA. Association between
socioeconomic status and renal function in a population of German
patients with diabetic nephropathy treated at a tertiary centre. Nephrolo
Dial Transpl 2011; 26: 40174023.
Kidney International (2015) 87, 784791

clinical investigation

D Dunkler et al.: Lifestyle/social factors and kidney disease in diabetes

20.
21.

22.

23.

24.

25.
26.

The Look AHEAD Research Group. Cardiovascular effects of intensive lifestyle

intervention in type 2 diabetes. New Engl J Med 2013; 369: 145154.
Gamble JM, Hoang H, Eurich DT et al. Patient-level evaluation of
community-based, multifactorial intervention to prevent diabetic
nephropathy in northern Alberta, Canada. J Prim Care Community Health
2013; 3: 111119.
Gong Q, Gregg EW, Wang J et al. Long-term effects of a randomised trial
of a 6-year lifestyle intervention in impaired glucose tolerance on
diabetes-related microvascular complications: the China Da Qing
Diabetes Prevention Outcome Study. Diabetologia 2011; 54: 300307.
Robinson-Cohen C, Littman AJ, Duncan GE et al. Physical activity and
change in estimated GFR among persons with CKD. J Am Soc Nephrol
2014; 25: 399406.
Jardine MJ, Hata J, Woodward M et al. Prediction of kidney-related
outcomes in patients with type 2 diabetes. Am J Kidney Dis 2012; 60:
770778.
Ritz E, Orth SR. Nephropathy in patients with type 2 diabetes mellitus.
New Engl J Med 1999; 341: 11271133.
Teo KK, Yusuf S, Sleight P et al. Rationale, design, and baseline
characteristics of 2 large, simple, randomized trials evaluating telmisartan,
ramipril, and their combination in high-risk patients: the Ongoing
Telmisartan Alone and in Combination with Ramipril Global Endpoint
Trial/Telmisartan Randomized Assessment Study in ACE Intolerant
Subjects with Cardiovascular Disease (ONTARGET/TRANSCEND) trials.
Am Heart J 2004; 148: 5261.

Kidney International (2015) 87, 784791

27.

28.

29.

30.
31.

32.

33.

34.

The ONTARGET investigatorsYusuf S, Teo KK et al. Telmisartan, Ramipril,

or both in patients at high risk for vascular events. N Engl J Med 2008;
358: 15471559.
Mann JFE, Schmieder RE, McQueen M et al. Renal outcomes with
telmisartan, ramipril, or both, in people at high vascular risk (the
ONTARGET study): a multicentre, randomised, double-blind, controlled
trial. Lancet 2008; 372: 547553.
Schmieder RE, Mann JFE, Schumacher H et al. Changes in albuminuria
predict mortality and morbidity in patients with vascular disease. J Am
Soc Nephrol 2011; 22: 13531364.
Levey AS, Stevens LA, Schmid CH et al. A new equation to estimate
glomerular filtration rate. Ann Intern Med 2009; 150: 604612.
Rosengren A, Hawken S, Ounpuu S et al. Association of psychosocial risk
factors with risk of acute myocardial infarction in 11 119 cases and 13 648
controls from 52 countries (the INTERHEART study): case-control study.
Lancet 2011; 364: 953962.
Alcohol in Moderation. Sensible drinking guidelines. Internet 2012;
http://www.drinkingandyou.com/site/pdf/SENSIBLE%2520DRINKING.pdf,
Accessed 8 September 2014.
Royston P, Sauerbrei W. Multivariable Model-Building. A Pragmatic
Approach to Regression Analysis Based on Fractional Polynomials for
Modelling Continuous Variables. John Wiley & Sons, Ltd.: Chichester, UK,
2008.
R Core Team. R: A Language and Environment for Statistical Computing.
R Foundation for Statistical Computing: Vienna, Austria, 2013.

791