Normothermic regional perfusion of kidneys: a single non-retrieval centre experience
Supplementary data for presentation O02 at BTS 2019
Robert Pearson, Colin Geddes, Marc Clancy, John Asher
27 February 2019
Donor characteristics
Donor age
# A tibble: 4 x 9
TxType `Mean donor age` SD `Median donor a… IQR Min Max `Male %`
<chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
1 DBD 49.7 16.6 52.5 24 2 78 49.2
2 DCD 53.2 13.8 54 17.5 2 79 51.2
3 LD 48.2 11.7 49 15 20 78 50.4
4 NRP 46.9 15.8 42 25 19 70 62.1
# … with 1 more variable: n <int>
Histograms of donor age
ANOVA for donor age
Df Sum Sq Mean Sq F value Pr(>F)
TxType 3 3283 1094.4 5.136 0.00159 **
Residuals 888 189218 213.1
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
10 observations deleted due to missingness Tukey multiple comparisons of means
95% family-wise confidence level
Fit: aov(formula = Donor_age ~ TxType, data = df2)
$TxType
diff lwr upr p adj
DCD-DBD 3.423921 0.234279 6.613564 0.0297712
LD-DBD -1.511872 -4.526631 1.502888 0.5690378
NRP-DBD -2.862158 -10.095174 4.370858 0.7386428
LD-DCD -4.935793 -8.397867 -1.473719 0.0014667
NRP-DCD -6.286079 -13.716671 1.144513 0.1303075
NRP-LD -1.350286 -8.707504 6.006932 0.9651300
Results from donors aged 19-70
However, going back to the maximum and minimum donor ages in the table above, the NRP group ranged from 19 to 70, while there were no living donors under 20 but there were paediatric donors in the DBD and standard DCD groups, and donors over 70 in all groups except NRP. To get a meaningful comparison, the remaining results exclude donors outside the 19-70 age range of the NRP group.
Donor renal function
# A tibble: 4 x 6
TxType `Mean donor eGFR` SD Median IQR n
<chr> <dbl> <dbl> <dbl> <dbl> <int>
1 DBD 88.6 29.7 92.2 33.7 313
2 DCD 95.6 25 99.5 30.2 196
3 LD 100 2.2 99.9 2.2 3
4 NRP 99.2 23.9 106. 24.0 28The donor eGFR for living donors only includes data on three cases! This is mainly because donor creatinine for living donors is poorly recorded on SERPR as the all have isotopic GFR measurements. Living donor transplants are therefore excluded from further analysis of donor eGFR.
ANOVA for donor eGFR
Df Sum Sq Mean Sq F value Pr(>F)
TxType 2 7667 3834 4.977 0.00722 **
Residuals 534 411286 770
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 Tukey multiple comparisons of means
95% family-wise confidence level
Fit: aov(formula = Donor_last_eGFR ~ TxType, data = df2.egfr)
$TxType
diff lwr upr p adj
DCD-DBD 7.065582 1.124349 13.00681 0.0148390
NRP-DBD 10.605888 -2.260122 23.47190 0.1292510
NRP-DCD 3.540306 -9.637248 16.71786 0.8028885
Recipient characteristics
Recipient age
# A tibble: 4 x 8
TxType Mean SD Median IQR Min Max n
<chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <int>
1 DBD 47.8 14.9 50 20 4 77.4 337
2 DCD 53.6 11.2 54.4 14.5 23.3 83.4 200
3 LD 41.1 17.2 44.3 27 1.9 72.2 251
4 NRP 49.8 13.6 49.4 15.4 11 72.8 29# A tibble: 4 x 4
TxType Adult Paediatric n
<chr> <int> <int> <int>
1 DBD 319 18 337
2 DCD 200 0 200
3 LD 221 30 251
4 NRP 28 1 29
ANOVA for recipient age
Df Sum Sq Mean Sq F value Pr(>F)
TxType 3 17928 5976 27.06 <2e-16 ***
Residuals 813 179562 221
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 Tukey multiple comparisons of means
95% family-wise confidence level
Fit: aov(formula = Age ~ TxType, data = df2.ages)
$TxType
diff lwr upr p adj
DCD-DBD 5.835704 2.420734 9.250674 0.0000726
LD-DBD -6.694994 -9.884817 -3.505171 0.0000005
NRP-DBD 2.029165 -5.374663 9.432994 0.8949054
LD-DCD -12.530698 -16.157018 -8.904377 0.0000000
NRP-DCD -3.806538 -11.408637 3.795560 0.5701799
NRP-LD 8.724159 1.220504 16.227815 0.0150669
Ischaemic times
# A tibble: 4 x 8
TxType Mean SD Median IQR Min Max n
<chr> <time> <time> <time> <dbl> <time> <time> <int>
1 DBD 12:51.132013 04:13.381080 11:50 20580 03:06 23:59 303
2 DCD 11:31.206897 03:34.244868 11:04 15405 06:02 23:59 174
3 LD 03:52.113924 01:52.702736 03:24 5040 01:24 20:30 237
4 NRP 09:51.615385 03:13.644208 09:18 16095 04:26 16:51 26Try to ignore the tiny fractions of a second in the mean and SD!
ANOVA for cold ischaemic time
Df Sum Sq Mean Sq F value Pr(>F)
TxType 3 1.511e+11 5.037e+10 329.3 <2e-16 ***
Residuals 736 1.126e+11 1.530e+08
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Unfortunately, neither post-hoc tests for intergroup differences nor diagnostic plots are available for date time objects. A Student’s t test was done to compare CIT between NRP and standard DCD:
Welch Two Sample t-test
data: CIT by TxType
t = 2.4263 secs, df = 34.8, p-value = 0.02058 secs
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
982.0145 secs 11059.1686 secs
sample estimates:
Time differences in secs
mean in group DCD mean in group NRP
41515.21 35494.62 Effect of NRP on delayed graft function
Effect observed in all recipients
The first set of results is for all recipients, including those with donors who had acute kidney injury requring renal replacement therapy at the time of donation.
Incidence of delayed graft function
# A tibble: 4 x 6
TxType `Delayed graft function` `Immediate function` `DGF %` `IF %` n
<chr> <int> <int> <dbl> <dbl> <int>
1 DBD 66 271 19.6 80.4 337
2 DCD 70 130 35 65 200
3 LD 13 238 5.2 94.8 251
4 NRP 6 23 20.7 79.3 29
Fisher's Exact Test for Count Data
data: DGF.risk
p-value < 2.2e-16
alternative hypothesis: two.sidedThese data show a difference in rates of delayed graft function which was statistically significant. The rate of DGF in the NRP group (20.7%) was between the rates observed in DBD (19.6%) and standard DCD (35.0%).
Duration of delayed graft function
Duration of delayed graft function in days, including zero days for immediate function:
# A tibble: 4 x 7
TxType `Mean duration` SD `Median duration` `Max duration` IQR n
<chr> <dbl> <dbl> <dbl> <dbl> <dbl> <int>
1 DBD 1.9 5.5 0 29 0 337
2 DCD 3.4 6.5 0 29 3 200
3 LD 1 4.8 0 29 0 251
4 NRP 2.5 7 0 27 0 29
The Kruskal-Wallis test is used here as the data are clearly not normally distributed, so the parametric ANOVA test would be inappropriate:
Kruskal-Wallis rank sum test
data: DGF by TxType
Kruskal-Wallis chi-squared = 60.733, df = 3, p-value = 4.099e-13 Comparison Z P.unadj P.adj
1 DBD - DCD -4.3805231 1.183948e-05 3.551843e-05
2 DBD - LD 4.1602793 3.178587e-05 6.357174e-05
3 DCD - LD 7.7847264 6.986420e-15 4.191852e-14
4 DBD - NRP -0.1741985 8.617095e-01 8.617095e-01
5 DCD - NRP 1.7981378 7.215517e-02 8.658620e-02
6 LD - NRP -1.9404262 5.232791e-02 7.849187e-02Histogram of duration of DGF
It is notable that the results in the NRP group are being affected by a single outlier, and as the numbers in the NRP group (19) are much smaller than the other DCD transplants (196) in the same time period, that single outlier is having a disproportionate effect. Review of the series shows that outlier received a transplant from a donor with acute kidney injury, who was still on CVVH at the time of donation, and additional the recipient had a ureteric complication.
Results excluding donors with AKI
The data is a little flawed as it uses eGFR calculated from donor creatinine, and so a donor with AKI receiving renal replacement therapy will appear to have normal eGFR. There were three transplants in the series from donors on RRT at the time of donation, two in the NRP group and one DBD. The results were for delayed graft function were reanalysed after excluding all the donors on RRT at the time of donation.
Incidence of delayed graft function
# A tibble: 4 x 6
TxType `Delayed graft function` `Immediate function` `DGF %` `IF %` n
<fct> <int> <int> <dbl> <dbl> <int>
1 DBD 65 271 19.3 80.7 336
2 DCD 70 130 35 65 200
3 LD 13 238 5.2 94.8 251
4 NRP 4 23 14.8 85.2 27Fisher’s exact test (all groups)
Fisher's Exact Test for Count Data
data: DGF3.risk
p-value < 2.2e-16
alternative hypothesis: two.sidedFisher’s exact test (NRP vs DCD)
Fisher's Exact Test for Count Data
data: DGF3a.risk
p-value = 0.04751
alternative hypothesis: true odds ratio is not equal to 1
95 percent confidence interval:
0.9899915 12.6530440
sample estimates:
odds ratio
3.059973 Duration of delayed graft function
Duration of delayed graft function in days, excluding donors with RRT at the time of donation:
# A tibble: 4 x 7
TxType `Mean duration` SD `Median duration` `Max duration` IQR n
<fct> <dbl> <dbl> <dbl> <dbl> <dbl> <int>
1 DBD 1.9 5.6 0 29 0 336
2 DCD 3.4 6.5 0 29 3 200
3 LD 1 4.8 0 29 0 251
4 NRP 1.5 5.4 0 27 0 27
The ANOVA assumptions have not been met, so need to use a non-parametric alternative, the Kruskal-Wallis test:
Kruskal-Wallis rank sum test
data: DGF by TxType
Kruskal-Wallis chi-squared = 61.791, df = 3, p-value = 2.435e-13As the overall differences are statistically significant, the Dunns test is used to examine differences between pairs of groups
Comparison Z P.unadj P.adj
1 DBD - DCD -4.468727 7.868653e-06 2.360596e-05
2 DBD - LD 4.121265 3.767975e-05 7.535950e-05
3 DCD - LD 7.838105 4.573955e-15 2.744373e-14
4 DBD - NRP 0.590557 5.548173e-01 5.548173e-01
5 DCD - NRP 2.522716 1.164525e-02 1.746788e-02
6 LD - NRP -1.114359 2.651252e-01 3.181502e-01Essentially, DBD vs DCD and all combinations of living donor vs other are statistically significant, the difference between DBD and NRP is non-significant (p=0.555) while the difference between NRP and standard DCD is statistically significant (p=0.017).
Effect of NRP on early post-transplant function
eGFR at seven days
All transplants
# A tibble: 4 x 8
TxType Mean SD Median IQR Min Max n
<fct> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <int>
1 DBD 45.3 34.4 38 43.7 4.5 228. 279
2 DCD 18.6 16.9 11.5 19.0 3.7 107. 152
3 LD 74.2 43.1 67.1 46.3 4.7 262. 229
4 NRP 41.1 31.5 28.6 31.4 5.7 135. 24
ANOVA for eGFR at 7 days (all transplants)
Df Sum Sq Mean Sq F value Pr(>F)
TxType 3 291594 97198 80.88 <2e-16 ***
Residuals 680 817220 1202
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 Tukey multiple comparisons of means
95% family-wise confidence level
Fit: aov(formula = eGFR_7d ~ TxType, data = df2.7d)
$TxType
diff lwr upr p adj
DCD-DBD -26.753683 -35.754327 -17.75304 0.0000000
LD-DBD 28.890129 20.929069 36.85119 0.0000000
NRP-DBD -4.220789 -23.212864 14.77129 0.9403084
LD-DCD 55.643812 46.303061 64.98456 0.0000000
NRP-DCD 22.532895 2.922436 42.14335 0.0168017
NRP-LD -33.110917 -52.266516 -13.95532 0.0000588There is a statistically significant difference between NRP kidneys and standard DCD, but NRP kidneys have almost the same function as DBD.
Transplants excluding donors with RTT for AKI
# A tibble: 4 x 8
TxType Mean SD Median IQR Min Max n
<fct> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <int>
1 DBD 45.5 34.3 38.5 43.8 4.5 228. 278
2 DCD 18.6 16.9 11.5 19.0 3.7 107. 152
3 LD 74.2 43.1 67.1 46.3 4.7 262. 229
4 NRP 42.6 31.2 29.3 31.6 8.9 135. 23
ANOVA for eGFR at 7 days (excluding RRT donors)
Df Sum Sq Mean Sq F value Pr(>F)
TxType 3 291594 97198 80.88 <2e-16 ***
Residuals 680 817220 1202
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 Tukey multiple comparisons of means
95% family-wise confidence level
Fit: aov(formula = eGFR_7d ~ TxType, data = df2.7d)
$TxType
diff lwr upr p adj
DCD-DBD -26.753683 -35.754327 -17.75304 0.0000000
LD-DBD 28.890129 20.929069 36.85119 0.0000000
NRP-DBD -4.220789 -23.212864 14.77129 0.9403084
LD-DCD 55.643812 46.303061 64.98456 0.0000000
NRP-DCD 22.532895 2.922436 42.14335 0.0168017
NRP-LD -33.110917 -52.266516 -13.95532 0.0000588
eGFR at 14 days
All transplants
# A tibble: 4 x 8
TxType Mean SD Median IQR Min Max n
<fct> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <int>
1 DBD 51.5 30.9 45.6 34.7 4.2 178. 291
2 DCD 30.4 21.4 26.6 29.9 3.5 120. 168
3 LD 75.0 38.8 68.1 37.8 6.1 239. 227
4 NRP 55.0 29.2 52.2 37.1 7.7 138 25
ANOVA for eGFR at 14 days (all donors)
Df Sum Sq Mean Sq F value Pr(>F)
TxType 3 195434 65145 64.46 <2e-16 ***
Residuals 707 714539 1011
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 Tukey multiple comparisons of means
95% family-wise confidence level
Fit: aov(formula = eGFR_14d ~ TxType, data = df2.14d)
$TxType
diff lwr upr p adj
DCD-DBD -21.113930 -29.046449 -13.181411 0.0000000
LD-DBD 23.480328 16.230772 30.729883 0.0000000
NRP-DBD 3.566213 -13.495905 20.628331 0.9496639
LD-DCD 44.594257 36.262495 52.926020 0.0000000
NRP-DCD 24.680143 7.130838 42.229448 0.0017804
NRP-LD -19.914115 -37.165471 -2.662758 0.0161076
Transplants excluding donors with RTT for AKI
# A tibble: 4 x 8
TxType Mean SD Median IQR Min Max n
<fct> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <int>
1 DBD 51.6 30.9 45.7 34.6 4.2 178. 290
2 DCD 30.4 21.4 26.6 29.9 3.5 120. 168
3 LD 75.0 38.8 68.1 37.8 6.1 239. 227
4 NRP 56.5 28.8 52.7 37.5 7.7 138 24
ANOVA for eGFR at 14 days (excluding RRT donors)
Df Sum Sq Mean Sq F value Pr(>F)
TxType 3 195434 65145 64.46 <2e-16 ***
Residuals 707 714539 1011
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 Tukey multiple comparisons of means
95% family-wise confidence level
Fit: aov(formula = eGFR_14d ~ TxType, data = df2.14d)
$TxType
diff lwr upr p adj
DCD-DBD -21.113930 -29.046449 -13.181411 0.0000000
LD-DBD 23.480328 16.230772 30.729883 0.0000000
NRP-DBD 3.566213 -13.495905 20.628331 0.9496639
LD-DCD 44.594257 36.262495 52.926020 0.0000000
NRP-DCD 24.680143 7.130838 42.229448 0.0017804
NRP-LD -19.914115 -37.165471 -2.662758 0.0161076
Effect of NRP on medium term function
eGFR at one year post-transplant
ANOVA for eGFR at 1 year
Df Sum Sq Mean Sq F value Pr(>F)
TxType 3 45203 15068 21.65 2.16e-13 ***
Residuals 667 464120 696
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
146 observations deleted due to missingness Tukey multiple comparisons of means
95% family-wise confidence level
Fit: aov(formula = eGFR_1yr ~ TxType, data = df2)
$TxType
diff lwr upr p adj
DCD-DBD -9.215678 -15.935981 -2.495375 0.0024840
LD-DBD 12.625544 6.386151 18.864937 0.0000015
NRP-DBD 3.102482 -11.091607 17.296571 0.9429804
LD-DCD 21.841222 14.741820 28.940623 0.0000000
NRP-DCD 12.318160 -2.274429 26.910748 0.1315364
NRP-LD -9.523062 -23.900517 4.854392 0.3213401
eGFR at two years post-transplant
ANOVA for eGFR at 2 years
Df Sum Sq Mean Sq F value Pr(>F)
TxType 3 24785 8262 12.11 1.17e-07 ***
Residuals 493 336306 682
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
320 observations deleted due to missingness Tukey multiple comparisons of means
95% family-wise confidence level
Fit: aov(formula = eGFR_2yr ~ TxType, data = df2)
$TxType
diff lwr upr p adj
DCD-DBD -7.79627566 -15.4879113 -0.10464 0.0455706
LD-DBD 10.53954883 3.3539460 17.72515 0.0010044
NRP-DBD 10.47851942 -6.9949117 27.95195 0.4108895
LD-DCD 18.33582449 10.1636802 26.50797 0.0000001
NRP-DCD 18.27479508 0.3730675 36.17652 0.0433534
NRP-LD -0.06102941 -17.7512380 17.62918 0.9999997
eGFR at three years post-transplant
ANOVA for eGFR at 3 years
Df Sum Sq Mean Sq F value Pr(>F)
TxType 3 12279 4093 6.397 0.000313 ***
Residuals 354 226499 640
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
459 observations deleted due to missingness Tukey multiple comparisons of means
95% family-wise confidence level
Fit: aov(formula = eGFR_3yr ~ TxType, data = df2)
$TxType
diff lwr upr p adj
DCD-DBD -4.781767 -13.621219 4.057684 0.5024451
LD-DBD 9.920099 1.716064 18.124133 0.0104676
NRP-DBD 7.791871 -11.133530 26.717272 0.7124255
LD-DCD 14.701866 5.483857 23.919875 0.0002784
NRP-DCD 12.573639 -6.812860 31.960137 0.3389158
NRP-LD -2.128227 -21.233373 16.976918 0.9917054
Multivariate models of eGFR
Models use analysis of covariance (ANCOVA) to control for the potential confounding effects of donor age, cold ischaemic time and donor renal function.
eGFR at 7 days
Df Sum Sq Mean Sq F value Pr(>F)
Donor_age 1 18176 18176 30.893 4.85e-08 ***
CIT 1 1030 1030 1.751 0.186
Donor_last_eGFR 1 1567 1567 2.664 0.103
TxType2 3 70882 23627 40.159 < 2e-16 ***
Residuals 420 247106 588
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
390 observations deleted due to missingnessAnova Table (Type III tests)
Response: eGFR_7d
Sum Sq Df F value Pr(>F)
(Intercept) 24403 1 41.4769 3.268e-10 ***
Donor_age 10454 1 17.7680 3.057e-05 ***
CIT 1956 1 3.3247 0.068956 .
Donor_last_eGFR 6426 1 10.9226 0.001031 **
TxType2 70882 3 40.1587 < 2.2e-16 ***
Residuals 247106 420
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Simultaneous Tests for General Linear Hypotheses
Multiple Comparisons of Means: Tukey Contrasts
Fit: aov(formula = eGFR_7d ~ Donor_age + CIT + Donor_last_eGFR + TxType2,
data = df2)
Linear Hypotheses:
Estimate Std. Error t value Pr(>|t|)
DCD - DBD == 0 -24.442 2.580 -9.473 <0.001 ***
LD - DBD == 0 65.173 14.352 4.541 <0.001 ***
NRP - DBD == 0 -10.681 5.421 -1.970 0.1715
LD - DCD == 0 89.615 14.320 6.258 <0.001 ***
NRP - DCD == 0 13.761 5.483 2.510 0.0485 *
NRP - LD == 0 -75.854 14.999 -5.057 <0.001 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
(Adjusted p values reported -- single-step method)
Simultaneous Confidence Intervals
Multiple Comparisons of Means: Tukey Contrasts
Fit: aov(formula = eGFR_7d ~ Donor_age + CIT + Donor_last_eGFR + TxType2,
data = df2)
Quantile = 2.4969
95% family-wise confidence level
Linear Hypotheses:
Estimate lwr upr
DCD - DBD == 0 -24.44222 -30.88441 -18.00004
LD - DBD == 0 65.17286 29.33769 101.00803
NRP - DBD == 0 -10.68115 -24.21602 2.85372
LD - DCD == 0 89.61508 53.86046 125.36970
NRP - DCD == 0 13.76107 0.07059 27.45155
NRP - LD == 0 -75.85401 -113.30547 -38.40255eGFR at 14 days
Anova Table (Type III tests)
Response: eGFR_14d
Sum Sq Df F value Pr(>F)
(Intercept) 46083 1 83.8889 < 2.2e-16 ***
Donor_age 18371 1 33.4421 1.451e-08 ***
CIT 4859 1 8.8448 0.003112 **
Donor_last_eGFR 5645 1 10.2756 0.001453 **
TxType2 49275 3 29.9000 < 2.2e-16 ***
Residuals 226875 413
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Simultaneous Tests for General Linear Hypotheses
Multiple Comparisons of Means: Tukey Contrasts
Fit: aov(formula = eGFR_14d ~ Donor_age + CIT + Donor_last_eGFR +
TxType2, data = df2)
Linear Hypotheses:
Estimate Std. Error t value Pr(>|t|)
DCD - DBD == 0 -21.317 2.513 -8.484 < 0.001 ***
LD - DBD == 0 45.724 13.872 3.296 0.00449 **
NRP - DBD == 0 -6.513 5.242 -1.243 0.56157
LD - DCD == 0 67.041 13.841 4.844 < 0.001 ***
NRP - DCD == 0 14.804 5.305 2.791 0.02219 *
NRP - LD == 0 -52.237 14.495 -3.604 0.00143 **
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
(Adjusted p values reported -- single-step method)
Simultaneous Confidence Intervals
Multiple Comparisons of Means: Tukey Contrasts
Fit: aov(formula = eGFR_14d ~ Donor_age + CIT + Donor_last_eGFR +
TxType2, data = df2)
Quantile = 2.4963
95% family-wise confidence level
Linear Hypotheses:
Estimate lwr upr
DCD - DBD == 0 -21.3172 -27.5895 -15.0448
LD - DBD == 0 45.7236 11.0950 80.3523
NRP - DBD == 0 -6.5132 -19.5988 6.5724
LD - DCD == 0 67.0408 32.4894 101.5922
NRP - DCD == 0 14.8040 1.5619 28.0460
NRP - LD == 0 -52.2369 -88.4206 -16.0531eGFR at 1 year
Anova Table (Type III tests)
Response: eGFR_1yr
Sum Sq Df F value Pr(>F)
(Intercept) 49548 1 114.2391 < 2.2e-16 ***
Donor_age 24817 1 57.2193 4.466e-13 ***
CIT 943 1 2.1751 0.14128
Donor_baseline_eGFR 249 1 0.5740 0.44926
TxType 2569 2 2.9610 0.05323 .
Residuals 134455 310
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Effect of NRP on graft and patient survival
Patient survival curve
Call:
survdiff(formula = pt.survival ~ TxType, rho = 0)
N Observed Expected (O-E)^2/E (O-E)^2/V
TxType=DBD 320 27 21.09 1.653 2.782
TxType=DCD 200 17 13.29 1.033 1.389
TxType=DCD-NRP 28 1 1.98 0.489 0.509
TxType=LD 223 7 15.63 4.761 6.811
Chisq= 7.9 on 3 degrees of freedom, p= 0.05 Graft survival curve
Call:
survdiff(formula = graft.survival ~ TxType, rho = 0)
N Observed Expected (O-E)^2/E (O-E)^2/V
TxType=DBD 320 34 27.35 1.617 2.707
TxType=DCD 200 19 17.22 0.184 0.246
TxType=DCD-NRP 28 2 2.65 0.160 0.166
TxType=LD 223 13 20.78 2.911 4.198
Chisq= 4.9 on 3 degrees of freedom, p= 0.2 These data show there were no graft losses within 7 days for the NRP group, and only one graft loss within the first year (96.3% survival). In contrast 97.0% of living donor kidneys, 92.4% of DBD and 92.7% of DCD kidneys were still functioning at one year. These results are all censored for death with functioning transplant.
Cox regression for graft survival
Call:
coxph(formula = graft.survival ~ TxType, data = df4)
n= 771, number of events= 68
coef exp(coef) se(coef) z Pr(>|z|)
TxTypeDCD -0.1194 0.8875 0.2865 -0.417 0.6769
TxTypeDCD-NRP -0.5013 0.6058 0.7280 -0.689 0.4911
TxTypeLD -0.6876 0.5028 0.3262 -2.108 0.0351 *
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
exp(coef) exp(-coef) lower .95 upper .95
TxTypeDCD 0.8875 1.127 0.5062 1.5560
TxTypeDCD-NRP 0.6058 1.651 0.1454 2.5233
TxTypeLD 0.5028 1.989 0.2653 0.9529
Concordance= 0.58 (se = 0.035 )
Rsquare= 0.007 (max possible= 0.664 )
Likelihood ratio test= 5.23 on 3 df, p=0.2
Wald test = 4.72 on 3 df, p=0.2
Score (logrank) test = 4.88 on 3 df, p=0.2Cox regression for patient survival
Call:
coxph(formula = pt.survival ~ TxType, data = df4)
n= 771, number of events= 52
coef exp(coef) se(coef) z Pr(>|z|)
TxTypeDCD -0.000344 0.999656 0.309679 -0.001 0.9991
TxTypeDCD-NRP -0.934433 0.392809 1.018663 -0.917 0.3590
TxTypeLD -1.050278 0.349840 0.424189 -2.476 0.0133 *
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
exp(coef) exp(-coef) lower .95 upper .95
TxTypeDCD 0.9997 1.000 0.54482 1.8342
TxTypeDCD-NRP 0.3928 2.546 0.05334 2.8925
TxTypeLD 0.3498 2.858 0.15233 0.8034
Concordance= 0.582 (se = 0.04 )
Rsquare= 0.012 (max possible= 0.567 )
Likelihood ratio test= 9.08 on 3 df, p=0.03
Wald test = 7.27 on 3 df, p=0.06
Score (logrank) test = 7.94 on 3 df, p=0.05Both Cox models show transplant type alone is a poor predictor of both graft and patient survival, except that living donor kidneys are associated with better survival. Adding in recipient age (for patient survival) and donor age (for graft survival) may improve matters:
Call:
coxph(formula = pt.survival ~ TxType + Age, data = df4)
n= 771, number of events= 52
coef exp(coef) se(coef) z Pr(>|z|)
TxTypeDCD -0.20481 0.81480 0.31256 -0.655 0.512289
TxTypeDCD-NRP -0.94481 0.38875 1.01861 -0.928 0.353643
TxTypeLD -0.86104 0.42272 0.42641 -2.019 0.043459 *
Age 0.04867 1.04988 0.01283 3.794 0.000148 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
exp(coef) exp(-coef) lower .95 upper .95
TxTypeDCD 0.8148 1.2273 0.4416 1.503
TxTypeDCD-NRP 0.3888 2.5723 0.0528 2.862
TxTypeLD 0.4227 2.3656 0.1833 0.975
Age 1.0499 0.9525 1.0238 1.077
Concordance= 0.687 (se = 0.042 )
Rsquare= 0.032 (max possible= 0.567 )
Likelihood ratio test= 24.79 on 4 df, p=6e-05
Wald test = 21.13 on 4 df, p=3e-04
Score (logrank) test = 22.09 on 4 df, p=2e-04Call:
coxph(formula = graft.survival ~ TxType + Donor_age, data = df4)
n= 771, number of events= 68
coef exp(coef) se(coef) z Pr(>|z|)
TxTypeDCD -0.16316 0.84946 0.28705 -0.568 0.5698
TxTypeDCD-NRP -0.38359 0.68141 0.72944 -0.526 0.5990
TxTypeLD -0.64798 0.52310 0.32715 -1.981 0.0476 *
Donor_age 0.02105 1.02128 0.01025 2.054 0.0399 *
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
exp(coef) exp(-coef) lower .95 upper .95
TxTypeDCD 0.8495 1.1772 0.4840 1.4910
TxTypeDCD-NRP 0.6814 1.4675 0.1631 2.8465
TxTypeLD 0.5231 1.9117 0.2755 0.9932
Donor_age 1.0213 0.9792 1.0010 1.0420
Concordance= 0.628 (se = 0.038 )
Rsquare= 0.012 (max possible= 0.664 )
Likelihood ratio test= 9.69 on 4 df, p=0.05
Wald test = 9.04 on 4 df, p=0.06
Score (logrank) test = 9.27 on 4 df, p=0.05