Data extraction

The following variables were selected from the Liver database due to substantive interest. The national STAR files have more information about each variable

Full list of variables extracted

Variables (see UNOS for abbreviations): ABO, AGE, BMI_CALC, COD, COD_WL, COD2, COD3, COMPOSITE_DEATH_DATE, DAYSWAIT_CHRON, DEATH_DATE, DGN_TCR, DGN_TCR2, DIAG, DON_TY, EDUCATION, END_DATE, END_STAT, ETHCAT, EXC_DIAG_ID, EXC_HCC, FINAL_MELD_PELD_LAB_SCORE, FINAL_SERUM_CREAT, FUNC_STAT_TRR, GENDER, GRF_STAT, GSTATUS, HCC_DIAG, HCC_DIAGNOSIS_TCR, HCC_EVER_APPR, LOS, MELD_DIFF_REASON_CD, MELD_PELD_LAB_SCORE, MULTIORG, NUM_PREV_TX, PRI_PAYMENT_CTRY_TRR, PRI_PAYMENT_TRR,PRVTXDIF, PSTATUS, PTIME, REGION, REM_CD, TRR_ID_CODE, TX_DATE, TX_YEAR, TX_PROCEDUR_TY

Inclusion Criteria:

  • Transplants from 1/1/2005-12/31/2021
  • Age 18 and older
  • HCC identified by diagnosis code in the United Network for Organ Sharing (UNOS)

Exclusion criteria (any of the following):

  • Re-transplantation
  • Multiorgan transplant
  • Acute liver failure
  • Cholangiocarcinoma

Recoded Variables:

  • EDUCATION was re-coded as 1 = None - GED, 2 = College, 3 = Post Grad
  • ETHCAT was re-coded as 1 = White, 2 = non-Hispanic Black, 3 = Hispanic / Latino, 4 = Asian, 5 = Other
  • PRI_PAYMENT_TRR (insurance) re-coded as 1 = Private, 2 = Public, 3 = Other. Only Private v. Public was retained for analysis

Table 1

Demographic and Clincal Variables for Donor Types
Deceased
(N=22383)		 Living
(N=616)		 Overall
(N=22999)
Era
Pre 17848 (79.7%) 400 (64.9%) 18248 (79.3%)
Post 4535 (20.3%) 216 (35.1%) 4751 (20.7%)
RECIPIENT AGE (YRS)
Mean (SD) 59.9 (7.28) 59.0 (9.25) 59.9 (7.34)
Median [Min, Max] 60.0 [18.0, 82.0] 60.0 [18.0, 76.0] 60.0 [18.0, 82.0]
Race / Ethnicity
White 14702 (65.7%) 459 (74.5%) 15161 (65.9%)
Asian 1591 (7.1%) 37 (6.0%) 1628 (7.1%)
Hisp/Lat 3895 (17.4%) 92 (14.9%) 3987 (17.3%)
Non Hisp Black 1909 (8.5%) 20 (3.2%) 1929 (8.4%)
Other 286 (1.3%) 8 (1.3%) 294 (1.3%)
Calculated Recipient BMI
Mean (SD) 29.0 (5.35) 28.0 (5.04) 28.9 (5.34)
Median [Min, Max] 28.4 [15.2, 63.1] 27.4 [16.3, 44.4] 28.4 [15.2, 63.1]
Missing 1 (0.0%) 0 (0%) 1 (0.0%)
WL MELD/PELD Lab Score at Most Recent Time
Mean (SD) 15.0 (8.46) 13.6 (5.52) 15.0 (8.40)
Median [Min, Max] 13.0 [6.00, 68.0] 13.0 [6.00, 34.0] 13.0 [6.00, 68.0]
Missing 8 (0.0%) 1 (0.2%) 9 (0.0%)
WL SERUM CREATININE AT REMOVAL
Mean (SD) 1.08 (0.679) 0.932 (0.439) 1.07 (0.674)
Median [Min, Max] 0.900 [0, 13.8] 0.870 [0.280, 6.70] 0.900 [0, 13.8]
Missing 9 (0.0%) 1 (0.2%) 10 (0.0%)
Functional Status
Mean (SD) 62.7 (19.8) 68.9 (17.3) 62.9 (19.8)
Median [Min, Max] 70.0 [10.0, 100] 70.0 [10.0, 100] 70.0 [10.0, 100]
GENDER
Male 17491 (78.1%) 423 (68.7%) 17914 (77.9%)
Female 4892 (21.9%) 193 (31.3%) 5085 (22.1%)
Payor Type
Private 11202 (50.0%) 378 (61.4%) 11580 (50.4%)
Public 11181 (50.0%) 238 (38.6%) 11419 (49.7%)
Level of Education
None to GED 10802 (48.3%) 235 (38.1%) 11037 (48.0%)
College 8713 (38.9%) 260 (42.2%) 8973 (39.0%)
Post Grad 1439 (6.4%) 68 (11.0%) 1507 (6.6%)
Missing 1429 (6.4%) 53 (8.6%) 1482 (6.4%)
Region
Central 1529 (6.8%) 22 (3.6%) 1551 (6.7%)
Midwest 3413 (15.2%) 130 (21.1%) 3543 (15.4%)
Northeast 5124 (22.9%) 290 (47.1%) 5414 (23.5%)
South 7935 (35.5%) 92 (14.9%) 8027 (34.9%)
Western 4382 (19.6%) 82 (13.3%) 4464 (19.4%)

Interrupted Time Series aka segmented regression

The purpose of this analysis was to test whether the time-series trajectory changed in 2015 and 2019 for live donor patients and deceased donor patients separately.

Because counts may be correlated with prior year counts (autocorrelation) we also examined the autocorrelation of residuals in a basic time series and segmented regression for LD and DD transplants. In both cases (below) introducing autocorrelations in the model did not change results.

Table of transplant counts per year per donor type

Transplant Year Deceased Donor Live Donor
2005 633 18
2006 874 24
2007 927 29
2008 1223 22
2009 1257 18
2010 1256 37
2011 1404 22
2012 1489 32
2013 1436 32
2014 1520 30
2015 1587 34
2016 1507 44
2017 1644 36
2018 1661 42
2019 1489 64
2020 1293 54
2021 1128 66
2022 786 39

Living donor regression model

Regression Coefficients and 95% Confidence Intervals for The Living Donor Time-series Model
Variable Estimate Lower 95% CI Upper 95% CI
(Intercept) 21.0 12.13 29.87
time 1.2 -0.46 2.86
int1 3.2 -17.29 23.69
time_int1 0.4 -6.55 7.35
int2 30.1 7.73 52.47
time_int2 -7.9 -17.44 1.64

Residual and Autocorrelation plots for LDLT

Figure 1 - Living Donor Liver Transplant Rates in Patients with HCC MELD exception over Pre-MMT era 1, Pre-MMT era 2, and MMT eras

Note: straight line represents the null hypothesis that trajectories did not change with policy changes

## Warning in predict.lm(mod1, interval = "prediction"): predictions on current data refer to _future_ responses

Deceased donor regression model

Regression Coefficients and 95% Confidence Intervals for The Deceased Donor Time-series Model
Variable Estimate Lower 95% CI Upper 95% CI
(Intercept) 785.47 667.73 903.21
time 92.54 70.48 114.59
int1 -108.33 -380.46 163.80
time_int1 -56.64 -148.90 35.62
int2 88.90 -208.22 386.02
time_int2 -263.30 -389.99 -136.61

Residual and Autocorrelation plots for DDLT

Figure 2: Deceased Donor Liver Transplant Rates in Patients with HCC MELD exception over Pre-MMT era 1, Pre-MMT era 2, and MMT eras

## Warning in predict.lm(mod1.D, interval = "prediction"): predictions on current data refer to _future_ responses

Logistic Regression of Donor Type on covariates

Here we see that the AIC is minimized when allowing for an interaction between era and payment type, and era and region. Model results below.

Reference groups for predictors:

##                      Model      AIC
## 1             Main effects 4728.608
## 2               Era by Age 4730.565
## 3         Era by Ethnicity 4733.421
## 4               Era by BMI 4729.244
## 5        Era by Creatinine 4727.614
## 6            Era by Gender 4727.490
## 7             Era by Payer 4725.705
## 8        Era by Functional 4725.067
## 9         Era by Education 4730.046
## 10           Era by Region 4681.941
## 11 Era by Region and Payer 4678.870
Odds ratios and confidence intervals of having a Living Donor Transplant (See above for reference groups)
Variable OR Lower 95% CI Upper 95% CI
(Intercept) 0.08 0.03 0.22
AGE 0.98 0.97 0.99
Eth_NewAsian 0.46 0.31 0.67
Eth_NewHisp/Lat 0.97 0.75 1.24
Eth_NewNon Hisp Black 0.28 0.17 0.44
Eth_NewOther 0.88 0.34 1.84
BMI_CALC 0.96 0.94 0.97
MELD_PELD_LAB_SCORE 1.00 0.98 1.01
FINAL_SERUM_CREAT 0.80 0.64 1.00
GENDERFemale 1.74 1.43 2.10
Post2019Post 1.60 0.56 4.04
Payment_newPublic 0.58 0.46 0.73
FUNC_STAT_PERCENT 1.01 1.01 1.02
ED_newCollege 1.31 1.08 1.57
ED_newPost Grad 1.88 1.40 2.50
REG_newMidwest 3.07 1.82 5.54
REG_newNortheast 3.83 2.32 6.82
REG_newSouth 0.40 0.22 0.78
REG_newWestern 1.47 0.85 2.70
Post2019Post:Payment_newPublic 1.52 1.06 2.20
Post2019Post:REG_newMidwest 0.56 0.20 1.73
Post2019Post:REG_newNortheast 1.40 0.54 4.09
Post2019Post:REG_newSouth 4.63 1.66 14.38
Post2019Post:REG_newWestern 0.64 0.22 2.02

Likelihood-ratio tests of all model predictors plus interactions from final interaction model.

Below, we see that all variables except MELD PELD a significant factors in the model

## Analysis of Deviance Table (Type II tests)
## 
## Response: DON_TY
##                      LR Chisq Df Pr(>Chisq)    
## AGE                    10.834  1  0.0009963 ***
## Eth_New                53.460  4  6.827e-11 ***
## BMI_CALC               28.457  1  9.578e-08 ***
## MELD_PELD_LAB_SCORE     0.218  1  0.6405216    
## FINAL_SERUM_CREAT       4.013  1  0.0451628 *  
## GENDER                 30.351  1  3.604e-08 ***
## Post2019               87.615  1  < 2.2e-16 ***
## Payment_new            17.160  1  3.436e-05 ***
## FUNC_STAT_PERCENT      21.839  1  2.966e-06 ***
## ED_new                 19.325  2  6.363e-05 ***
## REG_new               223.406  4  < 2.2e-16 ***
## Post2019:Payment_new    5.071  1  0.0243270 *  
## Post2019:REG_new       54.835  4  3.519e-11 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1