Risk Factors Predicting Survival of Liver Transplantation

Adult

Risk Factors Predicting Survival of Liver Transplantation I.H. Matinlauri, M.M. Nurminen, K.A. Höckerstedt, and H.M. Isoniemi ABSTRACT Prognostic models were developed for analyzing graft survival in a single-center study consisting of all 388 adult liver transplantations performed during 20 years. Proportional hazard models and generalized linear models were used to assess which risk factors, related to donor and recipient characteristics as well as graft preservation and operation, had an effect on graft survival. The prognostic modeling evidenced favorable trends in graft survival time during the successive quinquennials 1982–1987, 1988 –1992, and 1993–1997, in comparison to the referent time period 1998 –2002. Significant predictors of graft survival time were donor’s age, recipient-donor gender compatibility, recipient’s blood group, intraoperative blood transfusion, size of the transplanted organ, and indication for transplantation. Conventional histocompatibility matching did not correlate with graft outcome.

T

HE CLINICAL significance of both immunologic and nonimmunologic risk factors in liver transplantation (LTx) are still controversial on LTx outcome. The impact of HLA matching or presensitization in LTx is not clear.1– 4 The liver is generally believed to be less susceptible than kidney for immunologic risk factors. In this article, we studied graft survival and related adverse outcome events. More specifically, we analysed the effects of HLA compatibility, panel-reactive antibodies (PRA) and positivity in pretransplantation lymphocytotoxic cross-match on graft survival. Further clinical characteristics regarding the recipient and the donor, such as age, indication for transplantation, gender and blood group compatibility, and recipient’s cytomegalovirus status also were studied in respect to graft survival. The effect of intraoperative blood loss and transfusions on survival after LTx has been assessed by several different centers, with controversial data concerning associations

between blood use and patient and graft survival.5–7 Therefore, the effects of operational aspects, such as cold ischemia time, anhepatic time, intraoperative bleeding, intraoperative blood transfusions, and graft size, on graft survival also were evaluated. From the Department of Tissue Typing (I.H.M), Finnish Red Cross, Blood Service; Department of Public Health (M.M.N.), University of Helsinki, and Department of Epidemiology and Biostatistics, Finnish Institute of Occupational Health; and Transplantation and Liver Surgery Unit (K.A.H., H.M.I.), Helsinki University Hospital, Helsinki, Finland. The study was supported by University Research Grant for Finnish Red Cross, Blood Service, and a special governmental grant for health sciences research. Address reprint requests to Helena M. Isoniemi, Transplantation and Liver Surgery Unit, Surgical Hospital, Helsinki University Hospital, Kasarmikatu 11, 00130 Helsinki, Finland. E-mail: [email protected]

© 2005 by Elsevier Inc. All rights reserved. 360 Park Avenue South, New York, NY 10010-1710

0041-1345/05/$–see front matter doi:10.1016/j.transproceed.2004.12.078

Transplantation Proceedings, 37, 1155–1160 (2005)

1155

1156

PATIENTS AND METHODS Patient Population Between 1982 and 2002, all 388 adult cadaveric LTx performed in our center were included in the study. Of the 388 transplantations, 352 received a first transplant, 31 a second transplant, and 5 a third transplant. In the successive quinquennials, 1982–1987, 1988 –1992, and 1993–1997, 1998 –2002, the number of LTx performed was 17, 79, 125, and 167, respectively. All recipients were followed up for at least 1 year, the mean (⫾ SD) follow-up time was 4.5 (⫾ 4.2) years (range, 1–17.7 years). A number of potential risk factors related to donor and recipient characteristics and transplantation procedures were examined. Descriptive statistics of the most important covariates associated with the first and second LTx are reported in Table 1. In presenting the results for numerical covariates, continuous distributions were categorized to ease the interpretation of results. For example, the effect of the preservation time (cold ischemia time [CIT]) on the risk of graft survival was not linear and in the choice between entering CIT as a quadratic or categorical covariate the latter seemed the more practical alternative from the medical point of view.

HLA Typing, PRA, and Cross-Match Tests HLA class I typing was performed using a standard lymphocytotoxicity test. HLA-DR typing was performed using the cytotoxicity method until 1998 and subsequently using the polymerase chain reaction-sequence-specific oligonucleotide. PRA screening was performed using the cytotoxicity method until 1998 and then using flowcytometric HLA class I antibody screening (Flow PRA I, One Lambda, Calif, United States). Panel reactivity of ⱖ10% was considered as consistent with the presence of alloantibodies. Pretransplantation cross-matching was performed using the Amos modification of complement-dependent lymphocytotoxicity.

Blood Transfusion Data Information on the blood administered during transplantation was obtained from patient records. Cell saver has been used during operations within the last 12 years. The patients were divided into 3 groups according to blood used ⬍10 U, between 10 –19 U and ⱖ20 U.

Immunosuppression and Rejection Initial immunosuppression was based on cyclosporine, a tapering dose of methylprednisolone, and azathioprine. In the later period, some patients received tacrolimus-based immunosuppression. Acute rejection, if suspected on clinical grounds, was confirmed with a core needle biopsy and treated with increased corticosteroid boluses or in mild cases by increasing the basal immunosuppression levels. Steroid-resistant rejection was treated with OKT3 monoclonal antibody.

Statistical Methods Before embarking on multiple regression modeling, we first performed preliminary testing of the examined risk factors of graft survival time by entering each risk factor as a univariate term in the model. To study the joint effect of multiple factors on graft survival, we used Cox proportional hazards regression with a robust variance for the time-dependent model. For handling ties, we used the Efron approximation.8 The assumption of proportional hazards of the Cox regression model over time was checked by examination of

MATINLAURI, NURMINEN, HÖCKERSTEDT ET AL the Schoenfeld residuals9 separately for each of the potential predictor variates. The check indicated no severe nonproportionality. The relationship of survival time to a continuous covariate was depicted as a scattergram with a fitted smooth regression function. The probability of graft survival was plotted as a step function of the time since transplantation. All computations were carried out using the S-Plus system.10

RESULTS Univariate Analysis of Graft Survival

The analysis of the transplantation data was performed in 2 stages, beginning with univariate analysis of each potential risk factor. Graft survival was tested according to the number of HLA compatibilities and there were no significant differences between survival curves. The number of HLA mismatches had no significant effect on graft survival. PRA was positive for the patients receiving their first and second allografts in 18.2% and 9.7%, respectively. Pretransplantation cross-match was positive in 65 (18.8%) first and in 5 (16.7%) second transplant recipients. Pretransplantation alloimmunization did not have an effect on graft survival. A preliminary set of significant (at the probability level P ⫽ .10) covariates for graft survival time identified in the first step are shown in Table 2. The increase in the value of 3 continuous variates showed a decreasing trend in the graft survival time: donor’s age, total bleeding volume, and intraoperative blood transfusion. In addition, the following covariate categories were significant predictors: retransplantation, male recipient, male recipient–female donor incompatibility, donor’s blood group AB, blood transfusion ⬎20 U, reduced size of transplanted organ, and indication for transplantation due to cirrhosis, tumor, or retransplantation. These covariates were used as the starting set for the selection of the multivariate survival regression model. Cox Regression Analysis of Graft Survival

Some of the prognostic factors that were significant in the univariate analysis were no longer predictive for graft survival in the hazard regression modeling, whereas some new covariates became significant predictors. In the final model, we allowed separate baseline hazards for the year of transplantation periods because of unequal duration of follow-up. Because the observation times of graft survival are incomplete for the individuals who were alive at the end of the follow-up, we computed the predicted survival probabilities for the subgroups adjusted for the year of transplantation to allow unbiased comparison. The survivor curves refer to a patient in a calendar time period with average values of the model covariates for the whole data set. There was a consistent trend toward longer graft survival in the later periods. For example, the 2-year survival rates in the 4 successive 5-year calendar periods were 51 (confidence interval [CI], 29 – 89)%, 64 (CI, 52–77)%, 80 (CI, 74 – 87)%, and 90 (CI, 86 –94)%, respectively. The proportion of grafts that survived at least 5 years

RISK FACTORS PREDICTING SURVIVAL

1157

Table 1. Descriptive Statistics of Donor and Recipient Characteristics, Separately for the First and Second Transplantation Covariate

Recipient’s gender Donor’s gender Recipient’s age, y Indication for transplantation

Recipient’s blood group

Donor’s blood group

Blood group compatibility

HLA AB mismatches

HLA DR mismatches

PRA ⱖ10% Cross-match test Recipient’s cytomegalovirus status Cold ischemia time, h Anhepatic time, min Total bleeding, L Blood transfusion Graft size Recipient’s status at the end of follow-up

Recipient’s survival time, y Acute rejection Time to acute rejection, d

Category/Numerical

1st Tx n ⫽ 352

2nd Tx n ⫽ 31

Man Woman Man Woman Mean Range Acute liver failure Budd-Chiari syndrome Cirrhosis other* Immunologic disease† Liver tumor Other disorder‡ Retransplantation A AB B O A AB B O Compatible Identical Incompatible 0 1 2 3 4 0 1 2 Yes No Positive Negative Positive Negative Mean Range Mean Range Mean Range Mean Range Full Reduced Alive Retransplanted Dead Median 1st–3rd quartile Yes No Median 1st–3rd quartile

145 207 200 152 47.6 16.8–70.5 74 16 74 146 26 16 — 152 32 54 114 127 9 39 117 82 269 1 2 24 95 149 82 28 162 156 64 288 65 281 299 43 6.3 1.2–22.0 59 21–188 10.5 0–82 12.2 0–129 329 22 240 31 81 4 1.4–8 162 190 15 9–33

10 21 13 18 40.5 16.0–63.0 — — — — — — 36 11 5 3 12 10 3 2 16 5 25 1 — 4 7 14 6 1 12 16 3 28 5 25 29 1 5.8 2.3–11.5 56 33–85 7.5 2–27 11.6 0–30 29 2 18 5 8 4 0.8–8 10 21 21 10–81

*Alcoholic, cryptogenic, and posthepatitis cirrhosis. † Primary biliary cirrhosis (PBC), primary sclerosing cholangitis (PSC), and autoimmune (AI). ‡ Polycystic liver, metabolic illness, and Caroli syndromes.

1158

over the successive quinquennials 1987–1991, 1992–1996, and 1998 –2002 showed consistently more favorable trends compared with the referent time period 1982–1986. Hazard ratios (HR) were 0.84 (CI, 0.19 –1.79; P ⫽ .66), 0.50 (CI, 0.23–1.10; P ⫽ .094), and 0.22 (CI, 0.09 – 0.52; P ⬍ .001), respectively. Table 3 gives HR estimates of graft survival for the significant covariate (categories): donor’s age recipientdonor gender compatibility, recipient’s blood group, blood transfusion volume, size of a transplanted organ, and indication for transplantation. The likelihood ratio test (chi-square statistic ⫽ 54.8 on 20 df; P ⬍ .001) of the fitted model refers to the whole set of covariates. With advancing donor’s age the probability of graft survival decreased. The 8-year graft survival rates of the donors 16 –19, 20 –29, 30 –39, 40 – 49, and 50 –59 years were as follows: 83 (CI, 74 –94)%, 70 (CI, 56 – 87)%, 67 (CI, 57– 80)%, 73 (CI, 64 – 82)%, and 65 (CI, 54 – 80)%. For example, the risk of a shorter survival in the group that was 50 –59 years of age was 2.8-fold relative to that of the group that was 20 –29 years of age. The recipient-donor gender compatibility was most unfavorable when the recipient was male (M) and the donor was female (F) (HR ⫽ 1.9; CI, 1.0 –3.4); for example, the 10-year survival rate was as low as 42 (CI, 22– 80)%, whereas for the cases of MM, FF, and FM the survival rates were 56 (CI, 43–73)%, 73 (CI, 63– 83)%, and 77 (CI, 69 – 86)%, respectively. Blood group A had the best 10-year survival rate: 72 (CI, 63– 81)%. The blood groups AB, B, and O had poorer graft survival rates. The risk associated with blood group B relative to blood group A was significant (HR 2.0; CI, 1.1–3.5; P ⫽ .018), and that of blood group O (HR 1.6; CI, 1.0 –2.5; P ⫽ .04). In patients with ⬎20 U of blood transfusion, grafts were lost earlier compared with grafts in the ⬍10 U and 10 –19 U categories (HR ⫽ 1.6; CI, 1.0 –2.6; P ⫽ .067), the respective 5-year survival rates were 66 (CI, 54 –79)%, 78 (CI, 72– 85)%, and 77 (CI, 69 – 85)%. The indication for transplantation had a significant effect on graft survival. When cirrhosis was taken as the referent category, immunologic illness entailed the (statistically significantly) shortest graft survival time (HR ⫽ 0.52; CI, 0.29 – 0.92; P ⫽ .02). Liver tumor incurred an increased risk (HR ⫽ 1.8; CI, 0.95–3.5; P ⫽ .07). The following 8-year survival rates were estimated: tumor 39 (CI, 21–73)%, retransplantation 60 (CI, 45– 80)%, cirrhosis 64 (CI, 53–79)%, acute liver failure 70 (CI, 59 – 83)%, immunologic illness 80 (CI, 72– 87)%, and “other” condition 93 (CI, 83–100)%. Furthermore, a reduced size of the graft was found to be predictive of a reduced graft survival time compared with a full-sized graft (HR ⫽ 3.2; CI, 1.6 – 6.6; P ⫽ .0016). DISCUSSION

Overall graft survival of our patients was comparable with the European Liver Transplant Registry data from the same period.11 In our study, the observed 10-year probability was

MATINLAURI, NURMINEN, HÖCKERSTEDT ET AL

approximately 61%, whereas the univariate model with year of transplantation as the only predictor variate yielded the probability estimate of 64 (CI, 58 –71)%. The multivariate model estimated the 10-year probability to be 67 (CI, 61–75)%. The prognostic modelling evidenced favorable trends in graft survival time over the consecutive time periods. The same observation has been done in registry data from Europe11 and the United States.12 In clinical practice, it is not possible to allocate cadaveric liver allografts according to HLA match. In the present study, no statistically significant effect of HLA matching on graft survival was found, which is in accordance with other studies.4,13 Although better histocompatibility may be achieved in living related LTx, no beneficial effects of HLA matching in living donor liver transplants have been shown either.1,14 PRA positivity was found to be associated with crossmatch positivity in first LTx in our study, but neither of them had any association with graft survival. This is different from a study where a positive pretransplantation lymphocytotoxic cross-match significantly reduced 1-year graft survival.2 However, in the same study, HLA antibody positivity was not associated with survival. The discrepancy may be, at least partly, explained by the smaller sample size,

Table 2. Univariate P Values of Significance Tests for the Association of Covariates With Graft Survival Time Covariate

Two-sided P*

†

Transplantation order: 3rd-Tx Recipient’s gender:‡ Male Recipient-donor gender compatibility:§ Recipient ⫽ male, donor ⫽ female Donor’s age:储 30–39 y 40–49 y 50–59 y Donor’s blood group:¶ AB Total bleeding volume:# 12.2–82.0 L Blood transfusion:** 20 U Graft size:†† Reduced graft Indication for transplantation:‡‡ Immunologic disease (PBC, PSC, AI) Liver tumor Other disorder

.0056 .027 .073 .034 .076 .056 .048 .039 .0091 .032 .035 .060 .067

*z statistic (Wald’s test) for time variates in Tables 2 and 3. Relative to the following. † 1st Tx. ‡ Gender ⫽ female. § Recipient ⫽ female, donor ⫽ male. 储 Aged 20 –29 y. ¶ Blood group A. # Maximum 4.0 L. **10 –19 U. †† Full-sized graft. ‡‡ Cirrhosis (alcohol, cryptogenic, and postepatitis).

RISK FACTORS PREDICTING SURVIVAL

1159

more immunized patients, and more cross-match–positive patients in their study. The graft survival of male recipient-female donor gender mismatched cases was found to be significantly poorer compared with other combinations, which is in accordance with earlier studies.15 Poorer long-term graft survival of male recipients compared with female recipients has been shown in other studies too.7 Patients with end-stage liver cirrhosis with portal hypertension and low coagulation factors typically have disturbed hemostasis requiring largevolume transfusions of blood products during surgery. Large replacement of blood volume at the time of operation has been suggested to cause dilution of circulating antibodies and increase tolerance in LTx.2 Immune absorption and formation of immune complexes by soluble HLA antigens shed from the liver also has been proposed to explain better immunologic tolerance after LTx compared with kidney transplantation.16 Development of microchi-

merism and development of anti-idiotypic antibodies are frequently postulated explanations for increased tolerance after solid organ transplantation, mechanisms that are also associated with blood transfusions.17 On the other hand, a greater transfusion requirement during the transplantation also has been found to decrease survival7 as in our study. Lower survival in patients with high use of blood products depends on many factors because usually this group of patients has more serious liver disease. Other factors such as earlier abdominal operations also may affect bleeding tendency. There were significant differences in survival according to type of liver disease. The worst survival was in the category of liver tumors and the risk was significantly increased also in retransplantation. Our results differ from other studies in that the risk of graft loss in patients with acute liver failure was not increased compared with cirrhosis patients. In this study, we compared immunologic covarietes with

Table 3. Multivariate HR Estimates of Graft Survival Time Covariate

Donor’s age:* ⬍20 y 20–29 y 30–39 y 40–49 y 50–59 y 60 y and older Recipient-donor gender compatibility:† Recipient: female, donor: male Recipient: female, donor: female Recipient: male, donor: male Recipient: male, donor: female Recipient’s blood group:‡ A AB B O Blood transfusion volume:§ ⬍10 U 10–19 U ⱖ20 U Graft size:储 Full size Reduced size Indication for transplantation:¶ Acute liver failure Budd-Chiari Other cirrhosis Immunologic disease (PBC,PSC,AI) Retransplantation Liver tumor Other disorder *Aged 20 –29 y. † Recipient ⫽ female, donor ⫽ male. ‡ Blood group A. § Maximum 4 L. 储 Full-sized graft. ¶ Cirrhosis (alcohol, crytogenic, and posthepatitis.)

HR

95% Lower Confidence Limit

95% Upper Confidence Limit

2.33 1.00 2.20 1.93 2.77 2.53

1.06 Referent 1.10 0.98 1.33 0.84

5.13 category 4.40 3.80 5.75 7.58

.036

1.00 1.13 1.61 1.88

Referent 0.67 0.93 1.04

category 1.91 2.79 3.39

.64 .086 .036

1.00 1.25 1.98 1.61

Referent 0.72 1.13 1.02

category 2.17 3.48 2.53

1.04 1.00 1.59

0.65 Referent 0.97

1.65 category 2.62

1.00 3.21

Referent 1.55

category 6.63

0.86 0.98 1.00 0.52 1.30 1.81 0.35

0.43 0.32 Referent 0.29 0.63 0.95 0.08

1.72 3.0 category 0.92 1.72 3.45 1.46

Two-Sided P

.025 .059 .0062 .097

.42 .018 .040 .88 .067

.0016 .67 .97 .025 .48 .072 .15

1160

other donor and recipient factors as predictors of liver graft outcome. In the final model, recipient donor mismatch was the only immunologic predictor of the outcome. Furthermore, donor’s age, amount of intraoperative blood transfusion, recipients blood group, and indication for liver transplantation correlated with graft outcome. Graft survival was improved over time. REFERENCES 1. Takakura K, Kiuchi T, Kasahara M, et al: Clinical implications of flow cytometry crossmatch with T or B cells in living donor liver transplantation. Clin Transplant 15:309, 2001 2. Bishara A, Brautbar C, Eid A, et al: Is presensitization relevant to liver transplantation outcome? Human Immunol 63: 742, 2002 3. Suh K-S, Kim SB, Chang S-H, et al: Significance of positive cytotoxic cross-match in adult-to- adult living donor liver transplantation using small graft volume. Liver Transplant 8:1109, 2002 4. Neumann UP, Guckelberger O, Langrehr JM, et al: Impact of human leukocyte antigen matching in liver transplantation. Transplantation 75:132, 2003 5. Mor E, Jennings L, Gonwa TA, et al: The impact of operative bleeding on outcome in transplantation of the liver. Surg Gynecol Obstet 176:219, 1993 6. Koneru B, Harrison D, Rizwan M, et al: Blood transfusions in liver recipients. A conundrum or a clear benefit in the Cyclosporine/Tacrolimus era? Transplantation 63:1587, 1997

MATINLAURI, NURMINEN, HÖCKERSTEDT ET AL 7. Cacciarelli TV, Keeffe EB, Moore DH, et al: Effect of intraoperative blood transfusion on patient outcome in hepatic transplantation. Arch Surg 134:25, 1999 8. Cox DR, Oakes D: Analysis of Survival Data. London: Chapman and Hall; 1984 9. Venables WN, Ripley BD: Modern Applied Statistics With S-Plus, 3rd Ed. New York: Springer-Verlag; 1999 10. MathSoft Inc: S-PLUS 2000 Guide to Statistics. Seattle: Data Analysis Products Division, MathSoft Inc; 1999 11. European Liver Transplantation Registry: Available at: http//www.eltr.org.results. Accessed August 11, 2002 12. Roberts MS, Angus DC, Bryce CL, et al: Survival after liver transplantation in the United States: a disease specific analysis of the UNOS database Liver Transpl 10:886, 2004 13. Opelz G, Wujciak T, Döhler B, et al: HLA compatibility and organ transplant survival. Rev Immunogenetics 1:334, 1999 14. Kasahara M, Kiuchi T, Uryuhara K, et al: Role of HLA compatibility in pediatric living-related liver transplantation. Transplantation 74:1175, 2002 15. Brooks BK, Levy MF, Jennings LW, et al: Influence of donor and recipient gender on the outcome of liver transplantation. Transplantation 62:1784, 1996 16. Davies HS, Pollard SG, Calne RY: Soluble HLA antigens in the circulation of liver graft recipients. Transplantation 47:524, 1989 17. Dzik WH: Mononuclear cell microchimerism and the immunomodulatory effect of transfusion. Transfusion 34:1007, 1994