Transcript Impact of Dialysis Prescriptions and Practices on Outcomes

D
PPS
Dialysis Outcomes and Practice Patterns Study
Impact of Dialysis Prescriptions
and Practices on Outcomes
Friedrich K. Port, MD, MS
Arbor Research Collaborative for Health
Ann Arbor, Michigan
ESRD State of the Art, Boston, MA. April 23-26, 2009
DOPPS Overview
• Prospective observational study, 1997 – 2011
• Representative HD samples in 12 countries
• Practice-patterns in dialysis facilities and outcomes
• 4 Phases: consistent data collection internationally
• DOPPS 2: Added focus on incident HD patients
• DOPPS 3: Added processes of care and nutrition
• DOPPS 4*: Added practice trends and MD opinions
• Goal: Improve Outcomes in Hemodialysis:
- Mortality, Morbidity, and Quality of Life
* 2009-2011 funded by Amgen, Kyowa Hakko Kirin, and Genzyme
Dallas Conference, 1989
• Held PJ, Brunner F, Odaka M, Garcia J,
Port FK, Gaylin DS: Five-year survival for
ESRD patients in the U.S., Europe, and Japan
1982-87. Am J Kidney Dis 1990; 15: 451-457.
• US survival is lower than EDTA or Japan: Why
– US captures all deaths, other registries don’t (?)
– US case mix or practices explain the differences
– Mortality differs in the general populations
– Were authors simply wrong?
Outline
• International outcome comparisons
– Are outcomes differences real?
– Can we explain outcomes difference?
• Dialysis prescription over the last 20 years
• Opportunities to improve practices in dialysis
– Treatment time, at same Kt/V
– Blood pressure
– Phosphorus
• Focusing analyses on Practice Patterns may
improve evidence and have practical implications
5-Year Survival for ESRD Patients
Based on Registries, Adjusted for Diabetes and Age
60
54
48
40
5-Year
Survival
(%)
39
40
20
0
US
EDTA
US
Japan
Includes all dialysis and transplant patients Held et al. AJKD 15: 451, 1990
U 277 99
DOPPS I: Survival Among Hemodialysis Patients
in Japan, Europe, and the United States:
Adjusted for demographics
and comorbidities
Unadjusted
Survival (%)
100
100
Japan (ref)
90
Japan (ref)
90
Europe (RR=3.12)
Europe (RR=2.84)
80
80
US (RR=5.34)
70
70
60
60
50
50
0.0
1.0
2.0
Years
US/EU
RR =1.33
3.0
US (RR=3.78)
0.0
1.0
2.0
3.0
Years
DA Goodkin et al. JASN 14: 3270-3277, 2003
Mortality in the General Population
versus the Dialysis Patient Mortality
Nathan Levin’s Hypothesis:
In international comparisons, higher dialysis patient
mortality is partly explained by higher mortality in the
general population
Methods:
Correlate WHO data with Registry + DOPPS data
Yoshino et al JASN 2006, 17:3510-3519
Relationship of All-Cause Mortality Rates Between
Dialysis Patients (DP) and General Population (GP)
Unadjusted:
Yoshino et al JASN (2006)
Relationship of All-Cause Mortality Rates Between
Dialysis Patients (DP) and General Population (GP)
Adjusted for age in DP (overall median mean age [60.4 yr]) and
GP (overall median percentage of population aged 65 yr [15.8%]).
N=21 countries.
Yoshino et al JASN (2006)
International Differences
• Differences confirmed for US versus EU using
detailed adjustment for case mix and same data
collection for death ascertainment
• Better outcomes in Japan may be exaggerated since
selection to transplant of healthier patients is
minimal in Japan
• Background mortality partially explains differences*
• Question: Do practice differences contribute?
Outcomes by Vascular Access Use:
Problems with patient-based analyses
• Patients who use a catheter for dialysis tend to be
sicker patients
• Patients using a catheter have higher mortality than
patients using an AV fistula
• Is the higher mortality due to catheter use or due to
the selection of sicker patients?
The use of catheters varies widely from facility to
facility even when adjusted for case mix. This may
be a practice pattern: In DOPPS we studied overall
mortality in facilities by level of catheter use
Pisoni et al. AJKD 2009, 53: 475-491
Vascular Access Use and Mortality Risk
Facility-Based Model
RR of Death among Facility Patients
per 20% more facility use of indicated access
type 1.4
1.2
1.19
1.08
1.00
1
0.8
0.6
p<0.0001
p<0.0001
p=0.008
Ref.
Ref.
p<0.0001
Catheters
Grafts
Fistulae
*DOPPS I+II, 1996-2004; n=27,892; adjusted for age, gender, black race, yrs with ESRD, 14
comorbidity classes, weight, other practice indicators (median treatment time, % of pts with
S. Ca >10 mg/dl or S. PO4 >5.5 mg/dl) whether hosp unit, & accounted for facility clustering;
stratified by study phase & region. Facility access use is adjusted for facility case-mix.
Mortality Risk in Facilities that have
Greater Use of Catheters or AV Grafts versus low use
RR of death
Fac. Catheter Use
1.5
(R2=0.95)
1.45
1.31
1.26
1.24
1.25
1.14
1
1
1.07
1.38
Fac. Graft Use
(R2=0.966)
1.14
1
Quintiles for Graft
and Catheter Use
0.75
0.5
0
20
40
60
% Adjusted Facility Access Use
80
Mortality Risk for US versus European DOPPS is
Largely Explained by Vascular Access Practice
RR of Death
1.5
1.36
Adjusted for
Case Mix
1.00
+ Adjusted for Facility
Vascular Access Practice
1.06
1.00
1
p<0.0001
US
EUR
p=0.43
US
EUR
0.5
All models were adjusted for age, gender, race, time on dialysis, 14 summary comorbid
conditions, weight, unit type, facility median treatment time, facility % pts with serum
phos > 5.5 and serum Ca> 10 mg/dl, and stratified by study phase; accounted for facility
clustering effects. DOPPS I + II; n=24,398; *EUR=France, Germany, Italy, Spain, and UK.
International Differences
• Differences confirmed for US versus EU using
detailed adjustment for case mix and same data
collection for death ascertainment
• Better outcomes in Japan may be exaggerated since
selection of healthier patients to transplant is
minimal on Japan
• Background mortality partially explains differences
• Differences in vascular access practice explain most
of the mortality differences between Europe and US:
This points to an opportunity to improve vascular
access care and outcomes in the US
Dialysis Prescription
• Kt/V Trends
• Treatment time (TT) and mortality risk
(independent of Kt/V)
Mean Single-pool Kt/V in US HD Patients
during the Past 20 Years
sp Kt/V
1.8
Cross-sections of
patients by year
1.6
1.4
1.22
1.2
1.0
1.53
1.42 1.45
1.40
1.36
1.59 1.61 1.61
1.53
1.11
0.99
0.8
0.6
0.4
USRDS
CMMS
Special Studies
CMAS
DMMS
DOPPS 1
US DOPPS
DOPPS 2
DOPPS 3
0.2
0.0
1986
'88
'90
'92
'94
'96
'98
2000
'02
'04
'06
Year
Adapted from Port et al. CJASN 1:246-255, 2006
Mean and Median Patient Prescribed
Treatment Time in the US, by DOPPS Phase*
Minutes
230
225
225
221.5
225
222.8
220
215
211.7
210
210
205
Mean
Median
200
DOPPS 1
DOPPS 2
DOPPS 3
(n=3,856)
(n=2,260)
(n=1,814)
*Prevalent Cross-section of patients in each phase, weighted to
represent total facility sample size.
Distribution of Facility Treatment
Time by Country and Phase
Mean Facility Treatment Time (min)
350
325
300
275
250
225
200
175
Phase II III
ANZ
II III
II III
I II III
I II III
I II III
I II III
I II III
II III
I II III
I II III
Be
Ca
Fr
Ge
It
Ja
Sp
Sw
UK
US
Box-plots show the 25th to 75th percentiles (box) with median
(line) and 5th and 95th percentiles (whiskers)
Mortality Risk by Average Facility
Treatment Time as a Practice Pattern
RR Mortality*
1.4
1.2
RR=0.96 per 15 minutes, p=0.03
1.19
1.10
1.00
1.0
0.8
p=0.04
p=0.26
Ref.
< 210
210 - 240
240 +
0.6
Facility Average Treatment Time (minutes)
*Adjusted 2-stage model (instrumental variable)
Treatment Time and Mortality:
Summary
• Patients treated with longer dialysis sessions have
lower mortality risk at the same Kt/V (Saran et al 2008)
• Patients treated in dialysis facilities that use on
average longer treatment times have lower mortality
(this analysis focuses on the practice and minimizes
bias due to patient health status)
• The agreement of these results enhances the level of
evidence
Predialysis Systolic Blood
Pressure and Mortality Risk
A New Analytical Approach Using
Patient Exposure to Different Practices
— Prevalent HD Patients —
Facility Median Pre-dialysis SBP, by Region
Substantial Variation Between Facilities and Regions
Median Facility Pre-dialysis SBP
180
North America
170
Japan
EU & ANZ
160
150
140
25th = 142 mmHg
75th = 153 mmHg
130
25th = 131 mmHg
75th = 145 mmHg
120
25th = 147 mmHg
75th = 161 mmHg
110
100
N of facilities =
150
84
62
*Based on initial prevalent cross section patients (n=8000) with ESRD >3 months in
296 facilities in DOPPS III (2005-2008). SBP=systolic blood pressure
Within-Facility Distribution of Pre-Dialysis SBP*
Substantial Variation Across a Wide BP Range
% of Pts in Facilities
35
30
25
20
15
10
5
0
<100
100-110
110-120
120-130
130-140
140-150
150-160
160-170
170-180
≥180
Pre-Dialysis SBP (mmHg)
* 22,559 initial prevalent cross section patients with ESRD duration > 180 days
from 919 facilities in DOPPS I, II, III
Pre-HD Systolic BP and All-Cause Mortality
Facility Level BP
Patient Level BP
1.15
*
1.4
RR for an additional 10% of patients
compared to the ref category
RR of death§
*
RR of
death§
1.20
1.15
1.3
1.15
1.07 *
*
1.07 1.05 1.06*
1.06
1.03
1.2
1.01
1.10
1.01
1.1
1.01
1.05
1.0
1.00
Ref
0.99
*
0.8
1.00
1.00
0.9
0.95
0.97
0.89
0.7
0.6
*P<0.05
Ref
0.90
0.85
110 120 130
140 150 160
170 180
Pt achieved pre-dialysis SBP (mmHg)
110
120
130
140
150
160
170
180
Pre-dialysis SBP Group (mmHg)
§ 21,388 prevalent HD, 919 facilities Excludes patients with SBP <110 mm Hg. Cox models adjusted for age, gender, black
race, BMI, vintage, study phase, hemoglobin, s. albumin, phosphorus, creatinine, ferritin, PTH, intra-dialysis weight loss,
treatment time, catheter use, 13 comorbidities, stratified by country and accounted for facility clustering. Facility level
model also adjusted for facility mean levels of intra-dialysis weight loss, dialysate sodium, and treatment time (min), % of
catheter use and % pts in albumin, Hgb, Kt/V, and phosphate guidelines. No meaningful change with the addition of antihypertensive medications to the models, or with the addition of pts with SBP<110.
Facility Predialysis Systolic BP
and All-Cause Mortality
RR of death
1.12
Facility Level Mortality RR
per an additional 10% of patients by
category compared to the reference
1.08
1.06 *
1.05*
1.04
1
Ref
* p <0.05
0.96
110-129
130-160
Pre-dialysis SBP (mmHg)
>160
Predialysis Blood Pressure Levels
and Survival: Summary
• Optimal target BP has been difficult to identify, because
BP is influenced by health status
• Facility-based analyses provide insights by minimizing
bias due to patient health status, and by taking
advantage of the large between-facility variation in BP as
a likely reflection of practices or MD opinion
• Our data show that:
– Patients treated at facilities where more patients have
low pre-dialysis SBP (110-130 mmHg) have higher
mortality risk
– Patients treated at facilities where more patients have
high pre-dialysis SBP (>160 mmHg) have higher mortality
risk
Predialysis Blood Pressure Levels
and Survival: Conclusion
• These facility-level findings suggest that both
higher predialysis SBP (>160 mmHg) and lower
SBP (<130 mmHg) are associated with elevated
mortality risk
• The present results are not consistent with
KDOQI Guidelines (SBP <140 mmHg)
• A clinical trial is needed to identify optimal
predialysis SBP goals
Serum Phosphorus and
Mortality Risk
Patient-based analyses
and
Practice-based analyses
Practices of Better Control of High
Phosphorus and Mortality Risk
BACKGROUND:
• Patient level analyses showing higher mortality in
patients with high P levels may be confounded,
if sicker patients have higher P levels
• The new KDIGO Guidelines recommend control
“toward normal P levels” since randomized trials
are lacking
• Since randomization to poor P control is not
feasible, can we make observational studies more
informative?
Mortality Risk by Phosphorus Categories
Patient-Level Analyses, Among Patients on HD > 180 days
HR
4.0
All-cause
Cardiovascular
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0
0.5
1.5
2.5
3.5
4.5
5.5
6.5
7.5
8.5
9.5
10.5
mg/dl
Cox models used all DOPPS (n=25,529) and adjusted for age, sex, race, BMI, years on ESRD, 13
comorbid conditions, facility clustering. Hazard ratios and 95% confidence intervals (whiskers) for allcause (events n=5,857) and cardiovascular mortality (n events=1,930)
Tentori et al. AJKD 2008
Variation in Facility-Level
Serum Phosphorus
Facility % of Patients
N=899 facilities
60
50
40
30
20
10
0
<3.5
3.5-5
5-6
6-7
>7
Serum Phosphorus (mg/dl)
The % of patients having a serum PO4 of >7 mg/dl varies from
3% in some facilities to 40% of patients in other facilities.
Facility-Level Serum Phosphorus versus
All-Cause and CV Mortality Risks
Among Patients on HD > 180 days
HR associated with
10% more patients
in the phosphorus
category
1.3
All-Cause
Cardiovascular
1.2
+95% C.I.
1.1
Ref
1
0.9
0.8
≤3.5
3.6-5.0
5.1-6.0
6.1-7.0
>7.0
Phosphorus (mg/dl) category
Hazard ratios and 95% confidence intervals (whiskers) for all-cause (events n=5,857) and cardiovascular
mortality (events n=1,930). Models (n=20,561) were stratified by study phase and region and adjusted for facility
clustering effect; baseline patient age, sex, race, BMI, time on ESRD, 13 comorbid conditions, hemoglobin,
albumin, normalized protein catabolic rate, single-pool Kt/V, prior parathyroidectomy, and vitamin D prescription;
the percentage of patients at a facility with serum calcium <8.5, 8.6-10, and >10 mg/dL; and the percentage of
patients at a facility with serum PTH <100, 101-300, 301-600, and >600 pg/mL.
Tentori et al AJKD 2008
Can we use Principles of Randomization
in Observational Studies?
If patients are assigned randomly to facilities: Yes
• Instrumental variables may reduce treatment by
indication bias
• This is useful when large differences in practice
are observed
• In DOPPS, we use facility-level treatment
variables as instrumental variables
Facility-Level Treatment Variables
Rationale in DOPPS Design
• Patients usually select dialysis facilities by
factors independent of their own medical
condition – e.g. by proximity to home
• Average treatment patterns differ substantially
among facilities, in part due to provider opinion
or preferences
Facility-Level Treatment Variables
• Since variations in treatment preferences are
likely “random” with respect to medical
condition, this provides a “natural experiment”
with advantages similar to randomization in a
clinical trial
• Randomization provides balance across both
measured and unmeasured confounders
Facility-Level Treatment Variables as
Instrumental Variables: Caveats
• Other treatment practices may vary together
with the treatment of interest
– Action: We adjust also for other treatment
practices
• Unmeasured treatment practices may be
confounders
– Action: We measure many practices
Impact of Dialysis Prescriptions and
Practices on Outcomes: Summary
• The DOPPS approach has allowed identification of
opportunities to improve practices and outcomes, e.g.
– Treatment time (>4 hours thrice weekly)
– Systolic Blood pressure (130-160 mmHg pre-dialysis)
– Phosphorus (Avoid PO4 >6 mg/dl)
• International outcome differences are confirmed and the
US-Euro difference is largely explained by case-mix and
vascular access: Need to improve vascular access
• The instrumental variable approach is useful when
based on large differences in actual clinical practice
Acknowledgements
• Thanks to participating DOPPS facilities for their
data submission and dedication, and to patients
for completing questionnaires
• DOPPS is supported by scientific research grants
without restrictions on publications from
– Amgen (1996-2011)
– Kyowa Hakko Kirin (1999-2011 in Japan)
– Genzyme (2009-2011)
Tightness of Hgb Control*
and Mortality Risk
Among Facility Patients
* measured as facility standard deviation of Hgb levels
Facility Hgb Standard Deviation
(measure of “tightness of Hgb control” among facility patients)
Patients (%)
50
Facility with
Std Dev = 1.0 g/dl
40
30
Facility with
Std Dev = 3.0 g/dl
20
10
0
6
8
10
12
14
16
Hemoglobin (g/dl)
Some facilities may have larger variation (standard deviation) in patient
hemoglobin levels. This may be due to:
(1) greater comorbidity and variation in ESA-responsiveness
among patients in some facilities
(2) differences in facility practices that impact anemia control
Variation in Facility Hemoglobin
Standard Deviation
Facilities (%)
The mean facility Hgb level did
not significantly correlate with
the facility Std Deviation of Hgb
40
30
23.6
24.9
17.4
20
12.1
10
10.3
4.7
3.9
0.3
2.9
0
<0.6
0.6-0.8 0.8-1.0 1.0-1.2 1.2-1.4 1.4-1.6 1.6-1.8 1.8-2.0
>2.0
Facility Standard Deviation (g/dl) for
Hgb Levels among Facility Patients
n=921 facilities, DOPPS I, II, and III;
Facility Hgb Std Dev based upon a prevalent cross-section of study patients in a
facility at start of each DOPPS phase on dialysis > 180 days
Facility Std Deviation in Hemoglobin Levels
and Mortality Risk
Relative Risk
of death
RR= 1.10 per 0.5 unit higher Std Dev (p=0.001)
Adjusted for Facility Mean Hgb Levels
1.40
1.18
1.20
1.23
1.08
1.00
1.00
0.80
Ref
p=0.10
p=0.0003
p=0.003
0.60
< 1.1
1.1 - 1.35
1.35 - 1.7
Facility Std Dev for Hgb Levels
> 1.7
Adjusted for age, gender, black race, years with ESRD, body mass index, 14 comorbidity classes and
facility mean Hgb level; stratified by country and phase; accounted for facility clustering effects;
n=23,245, DOPPS 1 to 3. Facility hgb std dev based upon facility prevalent cross-section, pts on
dialysis > 180 days
Facility Std Dev in Hemoglobin Levels
and Mortality Risk
Relative Risk
RR= 1.11 per 0.5 unit higher Std Dev (p=0.002)
of death
Adjusted for Facility Mean Hgb Level
1.40
plus adjusted for 6 other facility practices
1.27
1.16
1.20
1.00
1.06
1.00
0.80
0.60
Ref
p=0.23
p=0.002
p=0.0008
< 1.1
1.1 - 1.35
1.35 - 1.7
> 1.7
Facility Std Dev (g/dl) for Hgb Levels
Adjusted for age, gender, black race, years with ESRD, body mass index, 14 comorbidity classes and
facility mean Hgb level and the facility practice of treatment time, Kt/V, catheter use, serum Ca and
PO4, and mean ESA dose; stratified by country and phase; accounted for facility clustering effects;
n=23,245, DOPPS 1 to 3. Facility hgb std dev based upon facility prevalent cross-section, pts on
dialysis > 180 days
Average Std Deviation in Hemoglobin
by Country & DOPPS Phase*
Ave of Facility
Std Dev, g/dl
1.8
1.7
Spain
1.6
UK
ANZ
1.5
Italy
1.4
Germany
France
1.3
1.2
UK
Canada
Sweden
Belgium
Japan
US
Japan
1.1
1
DOPPS I
DOPPS II
DOPPS III
*Baseline prevalent cross-section of patients on dialysis > 180 days for each
country & phase. Restricted to facility with at least 12 observations
Practices Associated with
Tighter Hgb Control at Facility Level
•
•
•
•
•
Having a narrower Hgb target range
Adjusting ESA dose more often (at least monthly)
Checking Hgb levels more often (at least weekly)
Prescribing ESAs for more patients (higher %)
Giving ESA i.v. rather than subcutaneously
Tighter Hgb Control at Facility Level*
Summary
Tighter Control of Hemoglobin at the Facility Level
• is associated with lower adjusted mortality at the facility
• is associated with certain practice patterns
• appears to be feasible according to the observed
improvements over time in most countries
* i.e. smaller standard deviation of Hgb across patients
Distribution of Facility Mean TT,
by Region and Phase
TT (Minutes)
300
280
*
*
*
*
DOPPS I
DOPPS II
DOPPS I
DOPPS II
#
260
240
220
200
180
Europe
Japan
* p <0.05 vs US of same phase
# p <0.05 vs US DOPPS I
DOPPS I
DOPPS II
US
n = 546 facilities
TT=Treatment Time
Saran et al. KI 69: 1222-8, 2006
Hospitalization Risk by Facility Catheter Use
Median Facility = 18% of Patients Use Catheters
Facility (%)
1.07
9 - 13
Infection-Related Hospitalization
1.03
13 - 18
RR=1.33 (p=0.008) per 20%
Higher Facility Catheter Use
1.28
18 - 28
1.56
28 +
1.24
9 - 13
Access-Related Hospitalization
1.18
13 - 18
RR=1.79 (p<0.0001) per 20%
Higher Facility Catheter Use
1.64
18 - 28
2.56
28 +
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
Relative Risk (95% CI)
Reference = AVF; adjusted for case mix and AV Graft use
Hospitalization Risk by Facility Graft Use
Median Facility = 32% of Patients Use Grafts
Facility (%)
Infection-Related Hospitalization
1.30
15 - 23
RR=1.11 (p=0.008) per 20%
Higher Facility Graft Use
1.28
23 - 32
1.33
32 - 50
1.51
50 +
Access-Related Hospitalization
1.38
RR=1.29 (p<0.0001) per 20%
Higher Facility Graft Use
15 - 23
1.61
23 - 32
1.75
32 - 50
2.32
50 +
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
Relative Risk (95% CI)
Reference = AVF; adjusted for case mix and Catheter use
Facility Target Pre-Dialysis Systolic BP
by Medical Director Survey
By Region
SBP Target
(mmHg)
180
North America
Japan
EU & ANZ
170
160
150
140
130
25th = 140 mmHg
75th = 160 mmHg
120
110
100
N of facilities
119
57
* DOPPS III Medical Director Survey. N = 236 facilities
60