title: timing of continuous renal replacement therapy and...

Supplemental Tables 1a, 1b and 1c: Indication1 for CRRT initiation Vs. Objective assessment of renal

dysfunction and fluid overload

Table 1a:

Acute Renal Failure1

(n = 145) Fluid Overload1

(n = 144) Both1

(n = 99)

Severe Renal Dysfunction2 140 (97%) 123 (85%) 95 (96%)

Worsening in Renal Dysfunction3 66 (46%) 76 (53%) 52 (53%)

Fluid Overload Severity ≤ 10%

10 – 20% > 20%

69 (48%) 32 (22%) 44 (30%)

56 (39%) 42 (29%) 46 (32%)

41 (41%) 24 (24%) 34 (35%)

1 Indications for CRRT initiation were subjectively determined based on the clinical judgment of the

attending physician at the time of CRRT initiation.

2 Injury or Failure categories of pRIFLE at the time of CRRT initiation

3 Worsening pRIFLE categorization from ICU admission to CRRT initiation

Table 1b:

Renal Dysfunction based on pRIFLE

No Renal Dysfunction based on pRIFLE

ARF is an indication for CRRT initiation

142 3

ARF is not the indication for CRRT

initiation 34 11

Sensitivity = 0.81 Specificity = 0.79 PPV = 0.98 NPV = 0.24

Table 1c:

Fluid Overload based on

Fluid Balance No Fluid Overload based

on Fluid Balance

Fluid Overload is an indication for CRRT

initiation 88 56

Fluid Overload is not an indication for CRRT

initiation 18 28

Sensitivity = 0.83 Specificity = 0.33 PPV = 0.61 NPV = 0.61

Title: Timing of Continuous Renal Replacement Therapy and Mortality in Critically Ill Children

Supplemental Data for the Manuscript

Supplemental Table 2a: Logistic Regression

Factor Univariable OR

(95% CI) Univariable

p-value Multivariable ORa

(95% CI) Multivariable

p-value

Timing of Initiation 1.06

(1.02, 1.13) 0.02

1.05 (1.01, 1.11)

0.03

Fluid Overload Category 1.79

(1.27, 2.56) 0.001

1.46 (0.95, 2.26)

0.09

Age Category 0.98

(0.66, 1.46) 0.92 NA NA

Study Period 0.71

(0.50, 0.99) 0.05

0.71 (0.47, 1.06)

0.09

Indication for CRRT 1.77

(1.24, 2.57) 0.002

1.84 (1.21, 2.86)

0.01

Sepsis 1.64

(0.92, 2.93) 0.09

1.17 (0.58, 2.34)

0.66

Oncologic Diagnosis 2.83

(1.42, 5.84) 0.004

3.04 (1.36, 7.06)

0.01

PIM 2 at ICU Admission 1.02

(1.00, 1.04) 0.03

1.02 (1.00, 1.05)

0.10

PELOD at CRRT Initiation

1.03 (1.00, 1.07)

0.08 1.02

(0.97, 1.07) 0.41

Shock 1.32

(0.70, 2.54) 0.39 NA NA

Respiratory Failure 1.42

(0.76, 2.66) 0.27 NA NA

Worsening in Renal Function

1.96 (1.10, 3.51)

0.02 1.13

(0.53, 2.43) 0.75

Severe Renal Failure at ICU Admission

0.51 (0.28, 0.91)

0.02 NA NA

Severe Renal Failure at CRRT Initiation

0.93 (0.46, 1.90)

0.85 NA NA

a Based on an initial multivariable logistic regression model including the following variables: Timing of

Initiation, Fluid Overload Category, Study Period, Indication for CRRT, Sepsis, Oncologic Diagnosis, PIM 2

score at ICU admission, PELOD score at CRRT Initiation and Worsening in Renal Function

NA Not application since the factor was not included in the logistic regression model

Supplemental Table 2b: Multi-variable Logistic Regression with Interaction term between Timing of

Initiation and Fluid Overload Category

Variable Adjusted Odds Ratio (95% Confidence Interval) p-value

Timing of Initiation 1.37 (1.11, 1.79) 0.01

Fluid Overload 2.59 (1.51, 4.58) < 0.001

PIM 2 at ICU Admission 1.02 (1.00, 1.05) 0.08

Presence of Oncologic Diagnosis 3.14 (1.44, 7.01) 0.01

Indication for CRRT 1.93 (1.27, 3.00) 0.003

Study Period 0.62 (0.40, 0.94) 0.03

Timing of Initiation X Fluid Overload 0.88 (0.80, 0.96) 0.01

Supplemental Table 2c: Multi-variable Logistic Regression with Timing of Initiation analyzed as a

categorical variable

Variable Adjusted Odds Ratio (95% Confidence Interval) p-value

Timing of Initiation 2.35 (1.38, 4.19) 0.01

Fluid Overload 3.51 (1.54, 8.32) < 0.001

PIM 2 at ICU Admission 1.02 (1.00, 1.05) 0.08

Presence of Oncologic Diagnosis 2.79 (1.29, 6.22) 0.01

Indication for CRRT 1.89 (1.25, 2.91) 0.003

Study Period 0.64 (0.42, 0.96) 0.03

Timing of Initiation X Fluid Overload 0.72 (0.56, 0.92) 0.01

Supplemental Table 3: Study Period, Timing of Initiation and Mortality

Study Period Outcome Category Timing of Initiation

Median (Inter-quartile Range)

Early (Jan 2000 – Dec 2002)

Mortality – 60%

All 4.0 (1.6, 8.4)

Survivors 2.7 (1.1, 7.1)

Non-Survivors 6.1 (2.5, 8.8)

Middle (Jan 2003 – Dec 2005)

Mortality – 43%

All 2.0 (0.7, 5.0)

Survivors 1.3 (0.5, 3.1)


Late (Jan 2006 – July 2009)

Mortality – 42%

All 2.2 (1.1, 4.2)

Survivors 2.1 (1.0, 3.3)


Supplemental Table 4: Stratified Cox Proportional Hazards Models

Variable Stratified based on

Study Period Hazard Ratio (p-value)

Stratified based on Age Group

Hazard Ratio (p-value)

Delayed CRRT Initiationb

5.95 (0.004) 5.95 (0.003)

Fluid Overloadc

2.83 (0.013) 2.83 (0.013)

Presence of Worsening in Renal Functiond 1.60 (0.050) 1.80 (0.020)

Presence of Oncologic Diagnosis 1.75 (0.021) 1.74 (0.024)

PELOD at CRRT Initiation

1.03 (0.017) 1.04 (0.005)

Supplemental Figure 1:

-1 0 1 2 3

05

10

15

20

Change in pRIFLE from ICU Admission to CRRT Initiation

Tim

e to

CR

RT

In

itia

tio

n (

da

ys)

Change pRIFLE N -1 5 0 95 1 31 2 31 3 28


1 2 3 4 5

05

10

15

20

Fluid Overload from ICU Admission to CRRT Initiation

Tim

e to

CR

RT

In

itia

tio

n (

da

ys)

FO N < 5% 52

5-10% 32 10-15% 24 15-20% 26 >20% 56

Supplemental Figure 3a:

0 5 10 15 20 25

020

40

60

Locally-weighted Polynomial Regression

Fluid Overload

Tim

e t

o C

RR

T I

nitia

tion

All Patients All Patients with correction for insensible losses Survivors with correction for insensible losses Non-survivors with correction for insensible losses

Supplemental Figure 3b:

0 5 10 15

05

10

15

20

Locally-weighted Polynomial Regression

Fluid Overload

Tim

e t

o C

RR

T I

nitia

tion

oooooo All Patients All Patients with correction for insensible losses Survivors with correction for insensible losses Non-survivors with correction for insensible losses


0 20 40 60 80 100 120 140

0.0

0.2

0.4

0.6

0.8

1.0

Unadjusted Survival Curves - Early vs. Late Initiators

Days after CRRT Initiation

Sur

viva

l Pro

babi

lity

Early Initiators

Late Initiators

Unadjusted HR 1.57 (p 0.038)

0 20 40 60 80 100 120 140

0.0

0.2

0.4

0.6

0.8

1.0

Adjusted Survival Curves - Early vs. Late Initiators


Sur

viva

l Pro

babi

lity

Early Initiators

Late Initiators

Adjusted HR 4.63 (p 0.009)

0 20 40 60 80 100 120 140

0.0

0.2

0.4

0.6

0.8

1.0

Unadjusted Survival Curves - Fluid Overload


Sur

viva

l Pro

babi

lity

FO < 10%

FO 10-20%

FO > 20%


0 20 40 60 80 100 120 140

0.0

0.2

0.4

0.6

0.8

1.0

Adjusted Survival Curves - Fluid Overload


Sur

viva

l Pro

babi

lity

FO < 10%

FO 10-20%

FO > 20%


0 20 40 60 80 100 120 140

0.0

0.2

0.4

0.6

0.8

1.0

Unadjusted Survival Curves - Worsening Renal Function


Su

rviv

al P

rob

ab

ility

No Change in Renal Function

Worsening Renal Function


0 20 40 60 80 100 120 140

0.0

0.2

0.4

0.6

0.8

1.0

Adjusted Survival Curves - Worsening Renal Function


Su

rviv

al P

rob

ab

ility

No Change in Renal Function

Worsening Renal Function


Supplemental Figure 5: Survival Curves – Early (Blue) Vs. Late (Red) CRRT Initiation

0 20 40 60 80 100 120 140

0.0

0.4

0.8


Surviv

orship

0 20 40 60 80 100 120 140

0.0

0.4

0.8


Surviv

orship

0 20 40 60 80 100 120 140

0.0

0.4

0.8


Surviv

orship

0 20 40 60 80 100 120 140

0.0

0.4

0.8


Surviv

orship

0 20 40 60 80 100 120 140

0.0

0.4

0.8


Surviv

orship

0 20 40 60 80 100 120 140

0.0

0.4

0.8


Surviv

orship

A. 2-day cut-off

Adjusted Hazard Ratio 2.05 (95% CI 0.62 – 6.73)

B. 3-day cut-off


F. 7-day cut-off


E. 6-day cut-off


C. 4-day cut-off


D. 5-day cut-off


Supplemental Figure 7: Fluid Overload in different pRIFLE categories

0 1 2 3 4

-20

020

40

60

Survivors

pRIFLE Category

Flu

id O

verload

0 1 2 3 4

-20

020

40

60

Non-Survivors

pRIFLE Category

Flu

id O

verload

0 1 2 3 4

-20

020

40

60

Early Initiators

pRIFLE Category

Flu

id O

verload

0 1 2 3

-20

020

40

60

Late Initiators

pRIFLE Category

Flu

id O

verload

Early Vs. Late Initiators - Propensity Score-based Matching Analysis

Methods:

Early and late initiator groups differed with respect to various baseline characteristics which could

potentially have confounding effect on the outcomes. Hence a propensity score-based matching was

performed. Propensity scores (PS) were used as a measure of conditional probability of early vs. late

initiation given the indication for CRRT initiation, pRIFLE category at ICU admission (as a measure of

severity of renal dysfunction) and PELOD prior to CRRT initiation (as a measure of severity of organ

dysfunction). PS was estimated using logistic regression model.

Matching was performed using the ‘Matching’ package in R (R version 3.0.1, The R foundation for

statistical computing). The cohort was divided into sub-groups based on multiple factors (PS quartiles,

Age group category, PIM 2 score quartiles, PELOD score quartiles and study epoch) and in each of the

sub-groups, the early (treatment) and late (control) initiators were matched 1:1 without replacement for

multiple covariates (PS, fluid overload category and presence of worsening renal function). The matched

groups were compared with respect to mortality. Also match balance was assessed comparing the two

groups with respect to various covariates using bootstrap Kolmogorov-Smirnov test.

Results:

Supplemental Table 5: Multivariable logistic regression model for Propensity Score estimation

Covariates in the Model Odds Ratio (95% Confidence Interval)

Severity of Renal Dysfunction at ICU Admission (pRIFLE category)

1.88 (1.40, 2.57)

Severity of Organ Dysfunction prior to CRRT Initiation (PELOD)

1.03 (0.98, 1.08)

Indication for CRRT Initiation (ARF or FO or Both)

1.09 (0.71, 1.67)

Matching based on propensity scores, fluid overload category and presence of worsening renal

dysfunction yielded 50 patients in each of the two groups – early and late initiators. Remaining patients

in the cohort were dropped due to absence of appropriate matches. A significant difference in mortality

was noted between the early and late initiators, with early initiators exhibiting 20% lower mortality

compared to the late initiators (44% vs. 64%, p-value 0.03). A good balance was noted between the two

groups, with respect to all other covariates (Supplemental Table 6). No significant differences with

respect to any of these covariates were noted between the two groups, after the propensity score-

based matching.

Supplemental Table 6: Early vs. Late Initiators before and after propensity score-based matching

Covariate % Difference in Mean between Early and Late Initiators

(Bootstrap Kolmogorov-Smirnov test p-value)

Before Matching After Matching

Propensity Scores 93% (< 0.001) 1% (0.63)

Fluid Overload -56% (0.003) -17% (0.54)

pRIFLE at Admission 87% (< 0.001) -8% (0.30)

pRIFLE at CRRT Initiation 18% (0.33) 11% (0.94)

Worsening in renal function from ICU Admission to CRRT Initiation

-90% (< 0.001) 18% (0.23)

PIM 2 scores -6% (0.26) -42% (0.49)

PELOD 50% (0.003) 2% (> 0.999)

Presence of Sepsis 8% (0.62) 24% (0.20)

Need for Vasopressor Support 56% (0.003) 38% (0.06)

Presence of Respiratory Failure 18% (0.28) 18% (0.32)

Presence of Oncologic Diagnosis -20% (0.25) -19% (0.39)

Indication for CRRT Initiation 17% (0.41) 15% (0.36)

title: timing of continuous renal replacement therapy and...

Documents