Title: Diapositiva 1
1 THREE AND TWELVE MONTH MORTALITY FOLLOWING FIRST
LIVER TRANSPLANTATION IN ADULTS IN EUROPE
PREDICTIVE MODELS FOR OUTCOME AK Burroughs, CA
Sabin, K Rolles, V Delvarto, V Karamo, R Adamo
and members of ELTAooLiver Transplantation and
Hepatobiliary Medicine Primary Care and
Population Sciences Royal Free Hospital and RFUC
Medical School, LONDON OELTR Hopital Paul
Brousse, PARIS ooAffiliated centres European
Liver Transplant Association (ELTA)
2ACKNOWLEDGEMENT
-
- The authors of Royal Free Hospital are grateful
to the ELTA committee for permission to use the
ELTR data
3BACKGROUND
- liver transplantation is a well established
surgery - measurement with publication of surgical outcomes
is increasingly used - - UK individual surgeons results of
coronary bypass surgery/aortic valve replacement
(Bridgewater BMJ 2005) - ELTR (Adam Lancet 2000) published intrinsic
mortality risk without identified risk factors.
However - - hazard ratios do not provide absolute
mortality rates - underlying hazard is not estimated
- - applied to whole time course of survival
- - included patients only up to 1997
- Current prognostic models poor (Jacobs 2005)
4AIM 1ST LIVER TRANSPLANTS IN ADULTS
- evaluate 3 month mortality as a measure of
- surgical mortality
- evaluate mortality at 1 year as a standard
- evaluation of surgical/medical
performance - apply statistical methods that allow
- estimation of absolute expected mortality
rates - according to identified risk factors
- - for an individual
- - for an individual centres
performance - include data up to 2003
5METHODS I
- ELTR database
- - 1968 - 23 European countries
- - recent audit confirmation of validity
(Karam 2003) - 52955 transplants from January 1988 to June 2003
- exclusions based on eligibility criteria
- - 7222 retransplants
- - 5486 children (lt15 years old)
- - 862 combined transplants
- - 55 no follow up information
- exclusions based on missing data
- - 3984 donor age
- - 4119 donor sex
- - 13 recipient sex
6METHODS II
- missing data associated with earlier transplants
- 1988-1991 - 27
- 1992-1995 - 21
- 1996-1999 - 5
- 2000-2003 - 5
- Evaluation of 3-month mortality 31094
transplants -
3678 deaths
- Evaluation of 12-month mortality 27165
transplants -
5031 deaths
7METHODS III
- Random splitting 3 month
12 month - training set (70) 21605
18852 - validation set (30) 9489
8313 - Factors associated with survival identified from
training set using multivariable logistic
regression - Adequacy final model assessed on validation set
by - - Comparison of scores in those alive/dead
(unpaired t-tests) - - Discrimination Harrells c index
- - Calibration Hosmer-Lemeshow statistic
8RECIPIENT CHARACTERISTICS IN TRAINING SET
9RECIPIENT CHARACTERISTICS IN TRAINING SET
10DONOR CHARACTERISTICS IN TRAINING SET
as
11Factors associated with 3-month survival
estimates from multivariable logistic regression
Year of transplant
1988-1991
1992-1995
1996-1999
2000-2003
Aetiology
Acute liver failure
HCC
Alcoholic cirrhosis
HCV cirrhosis
PBC
21605 liver transplants 2540 deaths
Other
Age of donor
lt40 years
41-60 years
gt60 years
Blood group
Identical
Compatible
Incompatible
Increased risk
Adjusted odds ratio
Decreased risk
12Factors associated with 3-month survival
estimates from multivariable logistic regression
HBVve
Recipient age (per 5 years older)
Split/reduced graft
UNOS status
1
2
3
4
21605 liver transplants 2540 deaths
Total ischemia time
gt13 hours
lt13 hours
Not known
Size of centre
lt36 /year
37-69 /year
gt70 /year
Adjusted odds ratio
Decreased risk
Increased risk
13FACTORS ASSOCIATED WITH SURVIVAL AT 3 12 MONTHS
favourable
adverse year of transplant gt2000
lt2000 aetiology
cirrhosis HCC acute failure donor
age 40
gt60 blood group match identical
compatible/incompat. recipient
younger
older graft type whole
split/reduced UNOS status
4
1, 2 total ischaemia (hr) lt13
13 centre sizeyear
70 36
all variables plt0.0001 and were the same for
both 3 and 12 mo mortality but had different
statistical weights
14RESULTS EXPECTED/OBSERVED DEATHS 3 MONTH
MORTALITY
21605 (training cohort) H-L statistic 8.94
(p0.35) c statistic0.691
2540 deaths
patients 2161 2161 2159
2165 2157 2160 2161
2160 2161 2160
9489 (validation cohort) H-L statistic 5.87
(0.83) c statistic0.688
1138 deaths
patients 949 948 950
948 949 949 949
949 949 949 Decile 1
2 3 4 5 6
7 8 9 10
15RESULTS EXPECTED/OBSERVED DEATHS 12 MONTH
MORTALITY
18852 (training cohort) H-L statistic 2.7
(p0.95) c statistic0.667
3391 deaths
patients 1884 1886 1886 1887
1881 1888 1885 1885
1885 1885
8313 (validation cohort) H-L statistic 27.54
(p0.002) c statistic0.662
1640 deaths
patients 831 831 832
831 831 832 831
832 831 831 Decile 1 2
3 4 5 6
7 8 9 10
16MAJOR CAUSES OF DEATH AT 3 AND 12 MONTHS
17Example 1 varying only donor age
18Example 2 varying ABO matching and donor age
19SUMMARY I
- data show importance of donor characteristics
- - recipient characteristics (e.g. MELD) are
insufficient - to predict outcome
- - gives basis for recipient/donor matching
- either based on each centre or
national system -
- models can be updated every 2-3 years
- - reference cohort should be the latest
cohort - - planning simple formula for ELTA website
- evaluation of new variables (e.g. renal
function) to - improve precision can be evaluated
statistically
20SUMMARY II
- ELTR database is a powerful tool for examining
- outcomes of liver transplantation
- models evaluated allow
- - estimation of mortality risk considering
- together both recipient characteristics
and donor - characteristics (potential and real)
- - any centre to evaluate performance and
- compare it to others over similar time
periods - 3-month mortality model is currently best as a
- reference standard for outcome, based on
evaluation - of both discrimination and calibration