Professional Documents
Culture Documents
http://www.kidney-international.org
& 2014 International Society of Nephrology
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
clinical investigation
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
ONTARGET n = 25,620
Death
n = 1028 (15%)
Sensitivity analyses
clinical investigation
Table 1 (Continued )
Characteristics at baseline
No.a
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)
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
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
(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?
clinical investigation
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
32.02% 31.46%
Frequency
Relative odds of
renal endpoint
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.
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
0.939 (0.8950.985)
P-Values
788
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
clinical investigation
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
REFERENCES
1.
DISCLOSURE
2.
ACKNOWLEDGMENTS
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.
clinical investigation
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
791