1. No category

clinical investigation
http://www.kidney-international.org
& 2013 International Society of Nephrology
see commentary on page 550
Change in the estimated glomerular filtration rate
over time and risk of all-cause mortality
Tanvir C. Turin1, Josef Coresh2, Marcello Tonelli3, Paul E. Stevens4, Paul E. de Jong5,
Christopher K.T. Farmer4, Kunihiro Matsushita2 and Brenda R. Hemmelgarn1,6
1
Department of Medicine, University of Calgary, Calgary, Alberta, Canada; 2Department of Epidemiology, Johns Hopkins Bloomberg
School of Public Health, Baltimore, Maryland, USA; 3Department of Medicine, University of Alberta, Edmonton, Alberta, Canada;
4
Kent Kidney Care Centre, East Kent Hospitals University NHS Foundation Trust, Canterbury, Kent, UK; 5Department of Nephrology,
University Medical Center Groningen, University of Groningen, Groningen, the Netherlands and 6Department of Community Health
Sciences, University of Calgary, Calgary, Alberta, Canada
Using a community-based cohort we studied the association
between changes in the estimated glomerular filtration rate
(eGFR) over time and the risk of all-cause mortality. We
identified 529,312 adults who had at least three outpatient
eGFR measurements over a 4-year period from a provincial
laboratory repository in Alberta, Canada. Two indices of
change in eGFR were evaluated: the absolute annual rate
of change (in ml/min per 1.73 m2 per year) and the annual
percentage change (percent/year). The adjusted mortality
risk associated with each category of change in eGFR was
assessed, using stable eGFR (no change) as the reference.
Over a median follow-up of 2.5 years there were 32,372
deaths. Compared to the reference participants, those with
the greatest absolute annual decline less than or equal to
5 ml/min per 1.73 m2 per year had significantly increased
mortality (hazard ratio of 1.52) adjusted for covariates and
kidney function at baseline (last eGFR measurement).
Participants with the greatest increase in eGFR of 5 ml/min
per 1.73 m2 per year or more also had significantly increased
mortality (adjusted hazard ratio of 2.20). A similar pattern
was found when change in eGFR was quantified as an annual
percentage change. Thus, both declining and increasing
eGFR were independently associated with mortality and
underscore the importance of identifying change in eGFR
over time to improve mortality risk prediction.
Kidney International (2013) 83, 684–691; doi:10.1038/ki.2012.443;
published online 23 January 2013
KEYWORDS: chronic kidney disease; epidemiology and outcomes;
mortality risk
Correspondence: Brenda R. Hemmelgarn, Division of Nephrology, Foothills
Medical Centre, 1403 29th Street NW, Calgary, Alberta, Canada T2N 2T9.
E-mail: Brenda.hemmelgarn@albertahealthservices.ca
Received 18 March 2012; revised 8 October 2012; accepted 18 October
2012; published online 23 January 2013
684
Studies have consistently demonstrated that more advanced
chronic kidney disease (CKD) is associated with an increased
risk of mortality across both general and high-risk populations.1–5 However, these reports have predominantly
considered kidney function at baseline, without consideration of how the change in kidney function over time
influences the risk of such outcomes. There has been a
growing interest in the association between change in kidney
function and risk of adverse outcomes. Although populationbased studies have reported an association between
declining kidney function specifically and adverse clinical
outcomes,6–11 kidney function can be highly variable and
improve over time in some patients.10,12 Although recent
studies have reported an association between improvements
in kidney function (increasing estimated glomerular filtration
rate (eGFR)) and risk of mortality,7,8 these studies were
limited by their select study population (CKD patients only7)
and small study size.7,8
Using a population-based cohort of individuals receiving
routine clinical care in a single Canadian province, we
investigated the association between changes in kidney
function over time and risk of all-cause mortality. We
explored change in kidney function using two indices:
absolute annual rate of change and the annual percentage
change. We hypothesized that both increasing and declining
eGFR would be associated with higher mortality risk, as
compared with stable kidney function.
RESULTS
Among the participants, 54.8% had an eGFR X90, 37.9%
had an eGFR in the range of 60–89, 4.9% had an eGFR in the
range of 45–59, 1.7% had an eGFR in the range of 30–44, and
0.7% had an eGFR in the range of 15–29 (all eGFR in ml/min
per 1.73 m2). The median number of measurements available
for the study participants was 3. The distribution of annual
rate of change appeared normal and centered near the
origin (Figure 1). The mean annual rate of change was
1.04 ml/min per 1.73 m2 per year (s.d.: 3.83), with a
median of 0.91 ml/min per 1.73 m2 per year (interquartile
Kidney International (2013) 83, 684–691
clinical investigation
TC Turin et al.: Short-term change in eGFR and ESRD
range (IQR): 2.98 to 1.07). The distribution of the annual
percentage change in eGFR, which also appeared normal,
is shown in Figure 1. The mean annual percent change in
eGFR was 1.52 percent/year (s.d.: 6.05), with a median of
1.07 percent/year (IQR: 3.77 to 1.34).
Compared with study participants, individuals excluded
because of an inadequate number of serum creatinine
measurements (less than three outpatient serum creatinine
measurements spanning a time period of four calendar
years—Figure 2) were younger, with fewer comorbidities and
0.15
Density
0.1
0.05
0
–20
–10 –5
0
5
10
20
Annual rate of change in eGFR
0.1
Density
0.08
0.06
0.04
0.02
0
–50
–25
–10 0 10
25
Annual percentage change in eGFR
50
Sensitivity analyses
Figure 1 | Distribution of annual rate of change and annual
percentage change in estimated glomerular filtration rate
(eGFR).
At least one
measurement
At least one
measurement
a higher level of eGFR at baseline (Supplementary Appendix
Table S1 online).
Among the study cohort, 135,804 (25.7%) had stable
kidney function (no change in kidney function over the
accrual period), 133,723 (25.6%) had a positive slope
(improved kidney function), and 257,785 (48.7%) had a
negative slope (declining kidney function). Participants
experiencing a greater annual decline or increase in eGFR
were more likely to be female and had a higher prevalence of
comorbidities, in comparison with those with stable kidney
function (Table 1).
Over a median follow-up of 2.5 years, there were 32,372
(6.1%) deaths. Adjusted mortality rates were higher, with
both declining and increasing eGFR (Table 2), as compared
with those with stable kidney function: the greater the change
in eGFR, the higher the mortality risk. Mortality rates
(per 1000 person-years) were highest for participants with an
increase in eGFR of 5 ml/min per 1.73 m2 per year or more
(rate 16.52; 95% confidence interval (CI): 15.78–17.25) and
participants with a decline in eGFR of 5 ml/min per 1.73 m2
per year or more (rate 11.27; 95% CI: 10.90–11.65). Similarly,
higher mortality rates were observed for increasing as well as
declining percentage change in eGFR (Table 3). The mortality
rate was highest (rate 15.15; 95% CI: 14.52–15.78) for
participants with an increase in eGFR of X7 percent/year or
more, followed by participants with a decline in eGFR
of X7 percent/year (rate 11.60; 95% CI: 11.20–11.99).
Compared with those with stable eGFR, the adjusted risk
of death was almost two-fold higher in participants with an
increase in eGFR of X5 ml/min per 1.73 m2 per year (hazard
ratio (HR) 2.20; 95% CI 2.10–2.31), whereas those with a
decline in eGFR of X5 ml/min per 1.73 m2 per year also had
2-fold increased risk (HR 1.52; 95% CI: 1.46–1.57) (Figure 3).
Similarly, we observed a U-shaped relation between percentage change in eGFR per year and all-cause mortality
(Figure 4). The risk of mortality was 2.02 times higher
(95% CI: 1.92–2.11) for the participants with an increase in
eGFR of X7 percent/year or more, and the mortality risk was
1.56 times higher (95% CI: 1.49–1.62) for the participants
with a decrease in eGFR of 7 percent/year or more.
When stratified by category of baseline kidney function
(eGFR X90, 60–89, 45–59, 30–44, and 15–29 ml/min/
1.73 m2), increasing as well as declining eGFR was associated
At least one
measurement
End of study
Year 1
Year 2
Year 3
Year 4
Follow-up for outcome ascertainment
after last measurement
31 March
2009
eGFR accrual during 1 May 2002 to 31 December 2007
Figure 2 | Overview of cohort creation. eGFR, estimated glomerular filtration rate.
Kidney International (2013) 83, 684–691
685
clinical investigation
TC Turin et al.: Short-term change in eGFR and ESRD
Table 1 | Baseline characteristics of study participants by annual absolute rate of change in eGFR
Annual absolute rate of change in eGFR (ml/min per 1.73 m2 per year)
p5
4
3
2
1
0
1
2
3
4
X5
62,402
(11.8)
28,457
(5.4)
40,676
(7.7)
55,649
(10.5)
70,601
(13.3)
135,804
(25.7)
44,542
(8.4)
31,612
(6.0)
20,861
(3.9)
13,857
(2.6)
24,851
(4.7)
Female gender
Aboriginal
Diabetes
Hypertension
58.6
(17.6)
63.2
2.8
23.2
54.8
59.2
(16.3)
62.0
2.2
18.7
50.3
59.6
(16.0)
60.4
2.1
18.1
59.8
59.6
(15.7)
59.8
1.9
17.3
48.8
59.7
(15.3)
59.3
1.8
17.5
48.1
60.0
(15.1)
57.9
1.9
18.0
48.7
60.1
(15.3)
56.3
2.0
18.4
50.0
59.5
(15.5)
57.7
1.9
18.8
49.4
58.8
(15.6)
58.7
2.0
19.0
48.9
58.3
(16.2)
60.5
2.3
19.6
49.0
55.9
(17.0)
63.1
3.0
19.3
46.3
Proteinuria
Normal
Mild
Heavy
Unmeasured
53.3
8.7
3.6
34.4
60.0
7.1
2.0
30.9
61.8
6.9
1.7
29.6
64.0
6.5
1.5
28.1
65.5
6.2
1.2
27.1
65.8
6.0
1.2
27.0
64.2
6.4
1.1
28.3
63.8
6.3
1.1
28.8
61.8
6.7
1.2
30.3
61.7
6.8
1.3
30.2
58.6
7.0
1.4
33.0
Kidney function at baseline
eGFR X90
eGFR 60–89
eGFR 45–59
eGFR 30–44
eGFR 15–29
14.9
51.2
18.1
10.3
4.1
21.4
56.4
13.5
6.1
2.0
25.4
54.8
12.4
5.2
1.7
32.4
50.4
11.1
4.4
1.4
39.2
46.3
9.4
3.7
1.3
41.7
45.9
8.3
3.1
0.9
39.1
50.0
7.8
2.7
0.4
40.8
50.4
6.5
2.1
0.2
44.1
48.5
5.8
1.5
0.2
47.8
45.6
5.0
1.5
0.1
57.0
38.9
3.5
0.6
0.0
Cerebrovascular disease
Peripheral vascular disease
CHF
COPD
Cancer
Myocardial infarction
Peptic ulcer disease
7.2
5.6
12.2
22.7
12.6
7.6
3.9
5.4
4.0
7.7
19.7
9.7
5.4
3.2
5.1
3.6
6.5
18.6
9.6
4.9
2.9
4.5
3.2
5.7
17.9
9.1
4.4
2.8
4.5
3.0
4.9
17.3
8.8
4.0
2.6
4.4
2.9
4.6
17.4
8.8
3.9
2.7
4.9
3.1
5.2
18.0
9.2
4.3
2.7
4.7
3.2
5.1
18.3
9.2
4.4
2.9
5.3
3.2
5.7
18.8
9.5
4.6
2.8
5.8
3.7
6.2
19.9
9.8
4.5
3.1
6.4
3.7
6.8
21.5
11.0
5.0
3.3
Socioeconomic status
Pensioner
Low
With subsidy
32.3
4.6
7.6
31.3
3.4
7.2
31.9
3.2
7.4
31.2
3.0
7.6
30.9
3.0
7.6
31.3
2.9
7.5
31.5
3.1
7.3
29.9
2.4
7.3
28.9
3.6
7.3
28.2
4.1
7.2
24.3
4.9
8.2
N (%)
Age, mean(s.d.), years
Abbreviations: CHF, congestive heart failure; COPD, chronic obstructive pulmonary disease; eGFR, estimated glomerular filtration rate.
Socioeconomic status was categorized as high income (annual adjusted taxable family income XCAD39250), low income (annual adjusted taxable family income oCAD
39250), low income with receiving social assistance, and pensioners (age X65 years).
Table 2 | Adjusted all-cause mortality rates, per 1000 person-years, by annual absolute rate of change in eGFR
Annual absolute rate of change in eGFR (ml/min per 1.73 m2 per year)
Events, n
Patients, n
p5
4
3
2
1
0
1
2
3
4
X5
6462
62,402
1808
28,457
2292
40,676
1853
55,649
3221
70,601
6426
135,804
2600
44,542
1793
31,612
1385
20,861
1088
13,857
2444
24,851
Adjusted rate
11.27
8.47
7.77
7.50
7.07
7.41
9.00
9.08
10.65
12.54
16.52
(95% CI)
(10.90–11.65) (8.05–8.90) (7.41–8.12) (7.19–7.81) (6.79–7.34) (7.19–7.64) (8.62–9.39) (8.62–9.53) (10.05–11.24) (11.76–13.33) (15.78–17.25)
Abbreviations: CI, confidence interval; eGFR, estimated glomerular filtration rate.
Rates are adjusted for age, sex, diabetes, hypertension, socioeconomic status, kidney function, proteinuria, and history of cancer, cerebrovascular disease, congestive heart
failure, chronic obstructive pulmonary disease, myocardial infarction, peptic ulcer disease, and peripheral vascular disease at baseline (last measurement).
with an increased risk of death across all categories of baseline kidney function (Supplementary Appendix Table S2 online
and Supplementary Appendix Figure S1 online). Similar
results were obtained when the change in eGFR was defined
as percentage change in eGFR per year (Supplementary
Appendix Table S3 online and Supplementary Appendix
686
Figure S2 online). Results were similar when baseline was
defined as the first serum creatinine measurement during the
accrual period. In addition, similar results were observed
when excluding participants with acute kidney injury–related
hospitalization during the eGFR accrual period, when
stratified analysis was carried out by baseline socioeconomic
Kidney International (2013) 83, 684–691
clinical investigation
TC Turin et al.: Short-term change in eGFR and ESRD
Table 3 | Adjusted all-cause mortality rates, per 1000 person-years, by annual percentage change in eGFR
Annual percentage change in eGFR (percent/year)
Events, n
Patients, n
Adjusted rate
(95% CI)
p7
6 to 5
4 to 3
2 to 1
0
1 to 2
3 to 4
5 to 6
X7
8665
57,111
2297
38,375
2932
65,837
4009
107,786
4187
113,847
3359
67,327
2319
37,845
1535
19,456
3069
21,728
11.60
(11.20–11.99)
8.70
(8.30–9.10)
7.74
(7.43–8.06)
7.41
(7.15–7.67)
7.43
(7.17–7.68)
9.03
(8.68–9.38)
9.87
(9.43–10.31)
11.26
(10.66–11.87)
15.15
(14.52–15.78)
Abbreviations: CI, confidence interval; eGFR, estimated glomerular filtration rate.
Rates are adjusted for age, sex, diabetes, hypertension, socioeconomic status, kidney function, proteinuria, and history of cancer, cerebrovascular disease, congestive heart
failure, chronic obstructive pulmonary disease, myocardial infarction, peptic ulcer disease, and peripheral vascular disease at baseline (last measurement).
10.00
Hazard ratio (95% CI)
2.20
1.68
1.52
1.14
1.05
1.01
1.00
1.22
1.22
1.43
0.95
1.00
(Reference)
0.10
–4
–3
–2
–1
0
1
2
3
4
.5
Proportion - –5
of patients (11.8%) (5.4%) (7.7%) (10.5%) (13.3%) (25.7%) (8.4%) (6.0%) (3.9%) (2.6%) (4.7%)
Annual rate of change in eGFR
Figure 3 | Risk of all-cause mortality by annual rate of change in estimated glomerular filtration rate (eGFR) adjusted for covariates at
the baseline (last measurement). Models were adjusted for age, sex, diabetes, hypertension, socioeconomic status, kidney function,
proteinuria, and history of comorbidities. CI, confidence interval.
Hazard ratio (95% CI)
10.0
2.02
1.56
1.17
1.04
1.22
1.32
1 to 2
(12.7%)
3 to 4
(7.1%)
1.51
1.00
1.0
0.1
Proportion
of patients
1.00
(Reference)
- –7
(10.8%)
–5 to –6 –3 to –4 –1 to –2
(7.2%) (12.4%) (20.4%)
0
(21.5%)
5 to 6
(3.7%)
.7
(4.1%)
Annual percentage change in eGFR
Figure 4 | Risk of all-cause mortality by annual percentage change in estimated glomerular filtration rate (eGFR) adjusted for
covariates at the baseline (last measurement). Models were adjusted for age, sex, diabetes, hypertension, socioeconomic status, kidney
function, proteinuria, and history of comorbidities. CI, confidence interval.
Kidney International (2013) 83, 684–691
687
clinical investigation
status, as well as when analysis was performed including
participants with an eGFR o15 ml/min/1.73 m2 (data not
shown).
DISCUSSION
In this community-based cohort, change in eGFR over a
period of up to 4 years (and requiring at least three eGFR
measurements) was associated with an independent and
graded increase in the risk of death. Compared with
participants with stable kidney function, both declining and
increasing eGFR was associated with a higher risk of death.
The risk was consistent across subgroups of kidney function
categories.
Our observations regarding declining eGFR and mortality
risk are consistent with other reports. The Cardiovascular
Health Study included 4380 community-dwelling older
adults with normal kidney function6 and reported a rate of
change of eGFR over a 7-year follow-up period. Sixteen
percent of participants experienced rapid decline in eGFR
(rate 43 ml/min per 1.73 m2 per year). Rapid decline in
kidney function was associated with a two-fold increased risk
of all-cause mortality. It is interesting to note that even in the
group of patients who had rapid decline in eGFR, the average
eGFR at the end of the follow-up period was 69 ml/min per
1.73 m2 (whereas average baseline eGFR for participants was
79 ml/min per 1.73 m2). Similarly, in our study, a change in
eGFR among higher levels of baseline kidney function was
also associated with increased risk of death, suggesting that
even with preserved kidney function the rate of change
has prognostic information for future mortality risk. The
Atherosclerosis Risk In Communities (ARIC) cohort8 also
examined the association between 3- and 9-year changes in
eGFR and the risk of death in 13,029 participants by dividing
the patients into quartiles on the basis of percentage annual
change in eGFR. The authors reported that the quartile of
patients with the greatest annual decline in eGFR over 3 years
was at a 22% greater risk of death compared with the patients
who experienced minimal annual decline in eGFR. Similarly,
patients with the greatest annual decline in eGFR over 9 years
were at 41% greater risk of death compared with the patients
who experienced minimal annual decline in eGFR. Data from
the Department of Veterans Affairs has provided similar
results9 where, during a median of 2.6 years, 10%, 28%, and
24% of participants experienced mild, moderate, and severe
CKD progression, respectively (defined as eGFR loss of 0–1,
1–4, 44 ml/min/year), with an increased risk of death for
those with moderate or severe CKD progression. Finally, a
recent study of 15,465 patients with stage 3 and 4 CKD
receiving primary care at a single institution7 reported an
84% increase in mortality for those with decreasing eGFR
(median –4.8 ml/min per 1.73 m2 per year, IQR: –8.2 to –3.2),
compared with those with stable eGFR.
Apart from eGFR, changes in serum albumin and
C-reactive protein (CRP) over time have also been studied
and were reported to be associated with adverse outcome
risk. Among adult hemodialysis patients, a decrease in serum
688
TC Turin et al.: Short-term change in eGFR and ESRD
albumin was associated with increased mortality risk.13
Although there was no association between change in
albumin and all-cause mortality in the Longitudinal Aging
Study Amsterdam,14 an increase in CRP levels15 was
associated with an increased mortality risk among the elderly.
Why is declining kidney function associated with an
increased risk for death? Declining kidney function may
contribute to increased risk by aggravating cardiovascular
risk factors, endothelial dysfunction, oxidative stress, and
vascular damage, as well as through activation of the
renin–angiotensin system induced by renal impairment.6,8,16,17 Further, worsening kidney function among
patients with relatively severe impaired kidney function
status may result in decreased appetite, decreased physical
function, and overall frailty,6,18,19 thus indirectly contributing
to a higher mortality risk among this subgroup.
Our finding that increasing eGFR is associated with excess
mortality is also similar to previous studies. Perkins et al.7
reported that, in comparison with patients with stable eGFR,
increasing eGFR was associated with a 42% increased risk of
death. Among ARIC study participants with stage 3 CKD,8
the group with minimal decline or an increase in eGFR
(annual change: –0.33 to 42.94%) also experienced a two-fold
increased risk of death. Recently Al-Aly et al.10 reported that
(compared with patients with mild CKD progression) those
who experienced no decline in kidney function exhibited
a trend toward increased risk of death (HR 1.15; 95% CI:
0.99–1.24). The explanation for the association between
increasing eGFR over time and increased risk of death is not
apparent, but this finding might be attributable to lower
serum creatinine generation as a result of reduced muscle
mass associated with chronic debilitating conditions.10,20
Although our analysis included only outpatient serum
creatinine measurements and further focused on those who
had measurements available over longer time horizons
(median eGFR accrual period 3.0 years) to minimize the
effect of acute kidney injury, the residual confounding from
resolving acute kidney injury may also have contributed to
the observed increased mortality risk associated with
improvement in kidney function. However, exclusion of
patients with acute kidney injury–related hospitalizations
during the eGFR accrual window did not qualitatively affect
our study results. These findings indicate toward the
possibility that increasing eGFR could be a marker of illness
rather than an independent risk factor for death.
Our study is strengthened by its large sample size, which
allowed us to study participants with a broad range of
baseline kidney function. Our study also has limitations. The
study cohort was limited to individuals who had outpatient
serum creatinine measurements as part of routine care, and
therefore does not include individuals who did not access
medical services. This might have resulted in inclusion of
patients with comorbid conditions associated with a more
rapid change in eGFR and increased risk of the adverse
outcomes. However, as we studied mortality among subjects
with an estimate of kidney function, this limitation does not
Kidney International (2013) 83, 684–691
TC Turin et al.: Short-term change in eGFR and ESRD
invalidate our findings. Given that we have used multiple
serum creatinine measurements over time, laboratory drift
over time may have influenced the study results. However,
the impact of this potential laboratory drift is expected to be
minimal, as we calibrated measurements across time periods
against a subset of healthy participants. Further, we have
previously reported on the increased mortality risk associated
with short-term changes in kidney function (adults with at
least two outpatient eGFR measurements during a 1-year
accrual period).21 The categorization of baseline kidney
function categories was based on eGFR values alone, which
may have led to misclassification of kidney function. In
addition, although we have adjusted for the presence and
severity of proteinuria, the majority of these measurements
were based on urinary dipstick, limiting our ability to assess
change in proteinuria levels over time. Finally, although we
adjusted for demographic factors, measured comorbidities,
and proteinuria, we were unable to adjust for covariates such
as body mass index, blood pressure control, cause of kidney
disease, and smoking status, introducing the possibility of
residual confounding. We also could not adjust for drug use,
as this information is available for a subsection of the
population in Alberta aged 65 years and older. However,
given the magnitude of the observed associations, this
limitation is unlikely to invalidate our conclusions.
In conclusion, we found that both declining and
increasing eGFR over time were independently associated
with mortality risk. These results suggest that monitoring
change in eGFR over time may enhance future mortality risk
prognostication in addition to the baseline kidney function.
MATERIALS AND METHODS
Study population and data source
The study cohort consisted of adults, aged 18 years or older, in
Alberta, Canada who had at least three outpatient serum creatinine
measurements spanning a time period of four calendar years
(Figure 2). We used the data repository of the Alberta Kidney
Disease Network22 to create the study cohort. The cohort accrual
period was from 1 May 2002 to 31 December 2007, with follow-up
extending to 31 March 2009 (the date up to which outcome data
were available). Patients receiving chronic dialysis or a kidney
transplant on or before cohort entry were identified from the
databases of Northern Alberta and Southern Alberta Renal Programs
and administrative data using a validated algorithm and were
excluded from the current analysis.23,24 Patients who developed endstage renal disease during the follow-up period were retained in the
analysis. Among 1,818,451 patients with at least one outpatient
serum creatinine measurement, there were 529,954 participants with
three or more measurements over four calendar years. After
exclusion of 642 participants with a first eGFR o15 ml/min per
1.73 m2, a total of 529,312 participants were included.
Magnitude of change in kidney function
The CKD-EPI equation25 was used to estimate the glomerular
filtration rate using outpatient serum creatinine measurements from
the accrual period. Serum creatinine measurements during the study
period were standardized to a central laboratory. This reference
laboratory (Capital Health Region, year 2009) used an isotope
Kidney International (2013) 83, 684–691
clinical investigation
dilution mass spectrometry reference standard. Gender-specific
correction factors were used to ensure province-wide
standardization of serum creatinine values over time. Change in
eGFR over time was estimated using all available outpatient eGFR
measurements for each patient during the accrual period. We used
two indices to describe the magnitude of change in eGFR: (a) the
absolute annual rate of change and (b) the annual percentage
change. The absolute annual rate of change in eGFR was calculated
by fitting a least-squares regression6 to all measurements for each
patient, where the slope of the regression line describes the absolute
rate of change for eGFR over time. The percentage change in eGFR
was calculated assuming a linear change on the log scale, consistent
with prior work.8 Given the size of the cohort, we were able to define
change in eGFR using a number of categories. The absolute annual
rate of change in eGFR was categorized as p 5, 4, 3, 2,
1, 0, 1, 2, 3, 4, andX5 ml/min per 1.73 m2 per year. The annual
percentage change in eGFR was categorized as p 7, 6 to 5,
4 to 3, 2 to 1, 0, 1–2, 3–4, 5–6, and X7 percent/year.
Assessment of covariates
Baseline was defined as the date of the last eGFR measurement
during the 4-year accrual period, and was the point at which followup for outcome ascertainment (all-cause mortality) commenced
(Figure 2). The date of the last eGFR measurement was chosen for
the baseline, as this is the point when the patient is seen by the
clinician and the time at which previous changes in kidney function
will be taken into consideration and extrapolated for prediction
of future risk. All covariates were assessed at the baseline. On the
basis of Government of Alberta health-care insurance records,26
socioeconomic status was characterized as high income (annual
adjusted taxable family income X$39,250 CAD), low income
(annual adjusted taxable family income o$39,250 CAD), low
income with subsidy (receiving social assistance), and pensioners
(65 years of age and older).3,22 Using validated algorithms27,28 from
hospital discharge records and physician claims, diabetes mellitus
and hypertension were identified. The Deyo classification of
Charlson comorbidities were identified from the physician claims
and hospitalization records using validated ICD-9-CM and ICD-10
coding algorithms.29 Kidney function at baseline was divided into
categories of eGFR X90, 60–89, 45–59, 30–44, and 15–29 ml/min per
1.73 m2, respectively. Baseline proteinuria was estimated by urine
albumin:creatinine ratio (ACR) or urine dipstick based on outpatient random spot urine measurements, and was categorized as
normal, mild, heavy, or unmeasured based on ACR (normal:
o30 mg/g, mild: 30–300 mg/g, or heavy: 4300 mg/g) or urine
dipstick (negative: no trace, mild: trace or 1 þ , or heavy: 2 þ ).3,30
Serum albumin or CRP levels were not available for the study
participants.
Assessment of outcome
The primary outcome of interest was all-cause mortality, as
determined from Vital Statistics data of the Alberta Health and
Wellness Registry file. Outcome ascertainment was prospectively
performed from the date of the last outpatient serum creatinine
measurement in the accrual period (baseline) to the end of the study
(31 March 2009).
Statistical analyses
Poisson regression was used to estimate all-cause mortality rates,
expressed per 1000 person-years of follow-up, for each group of
689
clinical investigation
TC Turin et al.: Short-term change in eGFR and ESRD
change in eGFR, after adjustment for sociodemographic variables,
baseline kidney function, proteinuria, and covariates. If the Poisson
assumption was not met, a quasi-Poisson model was used.31 Cox
proportional hazards models were used to estimate the adjusted risk
of all-cause mortality associated with each group of change in
kidney function, with stable kidney function (0 ml/min per 1.73 m2
per year for the absolute rate of change and 0 percent/year for
percentage change) used as the reference. The proportional hazards
assumption was tested and met. Participants were censored at
study end (31 March 2009) if they were still at risk or at an earlier
date if they experienced the event of interest or if they left the
province.
We performed several sensitivity analyses to verify the robustness
of our study findings. We repeated analyses stratified by baseline
eGFR category for rate of change by both absolute rate of change
and percentage change. We also repeated all analyses in which
‘baseline’ was defined as the first eGFR measurement during the 4year accrual period. We also repeated analyses excluding participants
who had an acute kidney injury–related hospitalization32 during the
eGFR accrual period. Analysis was also undertaken stratified by
socioeconomic status categories. Finally, we also repeated analyses
including participants with an eGFR o15 ml/min/1.73 m2.
Statistical analyses were performed using SAS version 9.2 (SAS
Institute, NC) and STATA version 11.2 (STATA, College Station,
TX). The institutional review board of the University of Calgary
approved the study.
14.
DISCLOSURE
15.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
All the authors declared no competing interests.
16.
ACKNOWLEDGMENTS
TCT is supported by Fellowship Awards from the Canadian
Institutes of Health Research, Canadian Diabetes Association, and the
Interdisciplinary Chronic Disease Collaboration (ICDC) team grant
funded by Alberta Innovates—Health Solutions (AI-HS). BRH and
MT are supported by AI-HS Salary Awards. BRH is supported by the
Roy and Vi Baay Chair in Kidney Research, and MT is supported by a
Canada Research Chair. JC and KM are supported by grants to the
CKD Prognosis Consortium from the National Kidney Foundation and
its sponsors.
17.
18.
19.
20.
SUPPLEMENTARY MATERIAL
Appendix Figure S1. Risk of all-cause mortality by annual rate of
change in eGFR across baseline levels of kidney function.
Appendix Figure S2. Risk of all-cause mortality by annual
percentage change in eGFR across baseline levels of kidney function.
Appendix Table S1. Characteristics of patients included and
excluded in the study cohort.
Appendix Table S2. All-cause mortality risk by annual rate of change
in eGFR, stratified by baseline eGFR category.
Appendix Table S3. All-cause mortality risk by annual percentage
change in eGFR, stratified by baseline (last measurement) eGFR
category.
Supplementary material is linked to the online version of the paper at
http://www.nature.com/ki
21.
22.
23.
24.
25.
26.
REFERENCES
1.
2.
690
Fried LF, Katz R, Sarnak MJ et al. Kidney function as a predictor
of noncardiovascular mortality. J Am Soc Nephrol 2005; 16:
3728–3735.
Go AS, Chertow GM, Fan D et al. Chronic kidney disease and the risks of
death, cardiovascular events, and hospitalization. N Engl J Med 2004; 351:
1296–1305.
27.
28.
Hemmelgarn BR, Manns BJ, Lloyd A et al. Relation between kidney
function, proteinuria, and adverse outcomes. JAMA 2010; 303:
423–429.
Turin TC, Tonelli M, Manns BJ et al. Chronic kidney disease and life
expectancy. Nephrol Dial Transplant 2012; 27: 3182–3186.
Chronic Kidney Disease Prognosis Consortium. Association of estimated
glomerular filtration rate and albuminuria with all-cause and
cardiovascular mortality in general population cohorts: a collaborative
meta-analysis. Lancet 2010; 375: 2073–2081.
Rifkin DE, Shlipak MG, Katz R et al. Rapid kidney function decline
and mortality risk in older adults. Arch Intern Med 2008; 168: 2212–2218.
Perkins R, Bucaloiu I, Kirchner H et al. GFR decline and mortality risk
among patients with chronic kidney disease. Clin J Am Soc Nephrol 2011;
6: 1879–1886.
Matsushita K, Selvin E, Bash LD et al. Change in estimated GFR associates
with coronary heart disease and mortality. J Am Soc Nephrol 2009; 20:
2617–2624.
Cheng TYD, Wen SF, Astor BC et al. Mortality risks for all causes and
cardiovascular diseases and reduced GFR in a middle-aged working
population in Taiwan. Am J Kidney Dis 2008; 52: 1051–1060.
Al-Aly Z, Zeringue A, Fu J et al. Rate of kidney function decline associates
with mortality. J Am Soc Nephrol 2010; 21: 1961–1069.
Shlipak MG, Katz R, Kestenbaum B et al. Rapid decline of kidney function
increases cardiovascular risk in the elderly. J Am Soc Nephrol 2009; 20:
2625–2630.
Shlipak MG, Katz R, Kestenbaum B et al. Rate of kidney function decline in
older adults: a comparison using creatinine and cystatin C. Am J Nephrol
2009; 30: 171–178.
Pifer TB, Mccullough KP, Port FK et al. Mortality risk in hemodialysis
patients and changes in nutritional indicators: DOPPS. Kidney Int 2002; 62:
2238–2245.
Schalk B, Visser M, Bremmer M et al. Change of serum albumin and risk of
cardiovascular disease and all-cause mortality. Am J Epidemiol 2006; 164:
969.
Alley DE, Crimmins E, Bandeen-Roche K et al. Three-year change in
inflammatory markers in elderly people and mortality: The Invecchiare in
Chianti Study. J Am Geriatr Soc 2007; 55: 1801–1807.
Manjunath G, Tighiouart H, Ibrahim H et al. Level of kidney function as a
risk factor for atherosclerotic cardiovascular outcomes in the community*
1. J Am Coll Cardiol 2003; 41: 47–55.
Weiner DE, Tighiouart H, Amin MG et al. Chronic kidney disease
as a risk factor for cardiovascular disease and all-cause mortality: a
pooled analysis of community-based studies. J Am Soc Nephrol 2004; 15:
1307–1315.
Shlipak MG, Stehman-Breen C, Fried LF et al. The presence of frailty in
elderly persons with chronic renal insufficiency. Am J Kidney Dis 2004; 43:
861–867.
Odden MC, Chertow GM, Fried LF et al. Cystatin C and measures
of physical function in elderly adults. Am J Epidemiol 2006; 164:
1180–1189.
Kovesdy CP, George SM, Anderson JE et al. Outcome predictability of
biomarkers of protein-energy wasting and inflammation in moderate
and advanced chronic kidney disease. Am J Clin Nutr 2009; 90:
407–414.
Turin TC, Coresh J, Tonelli M et al. One-year change in kidney
function is associated with an increased mortality risk. Am J Nephrol 2012;
36: 41–49.
Hemmelgarn BR, Clement F, Manns BJ et al. Overview of the Alberta
Kidney Disease Network. BMC Nephrol 2009; 10: 30.
Clement F, James M, Chin R et al. Validation of a case definition to define
chronic dialysis using outpatient administrative data. BMC Med Res
Methodol 2011; 11: 25.
Manns BJ, Mortis GP, Taub KJ et al. The Southern Alberta Renal Program
database: a prototype for patient management and research initiatives.
Clin Invest Med 2001; 24: 164–170.
Levey AS, Stevens LA, Schmid CH et al. A new equation to estimate
glomerular filtration rate. Ann Intern Med 2009; 150: 604–612.
Premium assistance program: Premium subsidy 2010; http://
www.health.alberta.ca/services/premium-subsidy.html (accessed
14 January 2013).
Hux JE, Ivis F, Flintoft V et al. Diabetes in Ontario: determination of
prevalence and incidence using a validated administrative data
algorithm. Diabetes Care 2002; 25: 512–516.
Quan H, Khan N, Hemmelgarn BR et al. Validation of a case definition
to define hypertension using administrative data. Hypertension 2009; 54:
1423–1428.
Kidney International (2013) 83, 684–691
clinical investigation
TC Turin et al.: Short-term change in eGFR and ESRD
29.
30.
Quan H, Sundararajan V, Halfon P et al. Coding algorithms for defining
comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care
2005; 43: 1130–1139.
Lamb EJ, MacKenzie F, Stevens PE. How should proteinuria
be detected and measured? Ann Clin Biochem 2009; 46:
205–217.
Kidney International (2013) 83, 684–691
31.
32.
Gardner W, Mulvey E, Shaw E. Regression analyses of counts and rates:
Poisson, overdispersed Poisson, and negative binomial models. Psychol
Bull 1995; 118: 392–404.
Waikar SS, Wald R, Chertow GM et al. Validity of international
classification of diseases, ninth revision, clinical modification codes for
acute renal failure. J Am Soc Nephrol 2006; 17: 1688–1694.
691