| |
HIV coinfection shortens the survival of patients with hepatitis C virus-related decompensated cirrhosis
|
| |
| |
Hepatology
April 2005
Juan A. Pineda 1
1Unidad de Enfermedades Infecciosas, Servicio de Medicina Interna, Hospital Universitario de Valme, Sevilla, Spain
"...The above results show that HIV coinfection shortens markedly the survival of patients with HCV-related cirrhosis after the first decompensation. In fact, roughly half of these patients die within a year of developing decompensated liver disease. This effect is independent of other predictors of survival, such as MELD and Child-Turcotte-Pugh class, age, HDV infection, the type of the first hepatic decompensation and the number of simultaneous complications....
...Our analysis was performed retrospectively. Ideally, the impact of HIV coinfection on the survival of patients harboring HCV-related cirrhosis should be assessed in a prospective study.
......In the multivariate analysis, HIV infection, a poorer Child-Turcotte-Pugh class, MELD score, more than one event at the first hepatic decompensation, older age, and seropositivity for anti-HDV at baseline - as well as the presence of HE (encephalopathy) as the first decompensation of cirrhosis - independently predicted survival...
....The data of this study indicate that the incidence of HE (hepatic encephalopathy) may be increased in HIV-coinfected patients with cirrhosis due to HCV. Likewise, the fact that HE is more frequent as the first decompensation suggests that it may occur earlier in HIV patients....
......The survival of HIV/HCV-coinfected patients was shorter than that of monoinfected subjects. The 1-year, 2-year, and 5-year survival estimates were 74%, 61%, and 44%, respectively, among individuals without HIV coinfection and 54%, 40%, and 25%, respectively, among coinfected patients. Other baseline factors that predicted survival in the univariate analysis were Child-Turcott-Pugh class, MELD score, older age, the event that defined the first decompensation episode, the number of simultaneous complications, and a positive result of serum HDV antibody testing....."
HIV infected were younger at death: median age 38 vs 68 yrs; and more likely to have higher daily alcohol intake.
Abstract
The impact of human immunodeficiency virus (HIV) coinfection on the survival of patients with hepatitis C virus (HCV)-related end-stage liver disease (ESLD) is unknown. Because HIV infection is no longer considered an absolute contraindication for liver transplantation in some countries, it has become a priority to address this topic.
The objective of this study was to compare the survival of HIV-infected and HIV-uninfected patients with decompensated cirrhosis due to HCV. In a retrospective cohort study, the survival of 1,037 HCV monoinfected and 180 HCV/HIV-coinfected patients with cirrhosis after the first hepatic decompensation was analyzed.
Of the group, 386 (37%) HCV-monoinfected and 100 (56%) HCV/HIV-coinfected subjects died during the follow-up. The median survival time of HIV-infected and HIV-uninfected patients was 16 and 48 months, respectively (P < .001).
The relative risk (95% CI) of death for HIV-infected patients was 2.26 (1.51-3.38). Other independent predictors of survival were age older than 63 years (2.25 [1.53-3.31]); Child-Turcotte-Pugh class B versus class A (1.95 [1.41-2.68]) and class C versus class A (2.78 [1.66-4.70]); hepatitis D virus infection (1.56 [1.12-4.77]); model for end-stage liver disease score, (1.05 [1.01-1-11]); more than one simultaneous decompensation (1.23 [1.12-3.33]); and the type of the first hepatic decompensation, with a poorer prognosis associated with encephalopathy compared with portal hypertensive gastrointestinal bleeding (2.03 [1.26-3.10]).
In conclusion, HIV coinfection reduces considerably the survival of patients with HCV-related ESLD independently of other markers of poor prognosis. This fact must be taken into account to establish the adequate timing of liver transplantation in HIV-coinfected subjects.
AUTHOR DISCUSSION
Our analysis was performed retrospectively. Ideally, the impact of HIV coinfection on the survival of patients harboring HCV-related cirrhosis should be assessed in a prospective study. Such a design would avoid the potential bias due to lack of uniformity in patient management. However, with the exception of liver transplantation, most therapeutic interventions in decompensated HCV-related cirrhosis lead to improvements in morbidity, whereas only some of them have a limited impact on mortality. Accordingly, we failed to find differences in survival related with the hospitals or the units the patients attended. Moreover, the present study allowed us to achieve a quicker response about the question analyzed than prospective surveys would have allowed. Prospective studies would need years to be completed. Because the survival of HIV-coinfected patients with decompensated HCV-related cirrhosis is short, a rapid response was required; consequently, the information provided by these results is relevant and useful at least until prospective studies are carried out. The large sample size is another strong point of this study, because this can reduce potential biases.
The mechanisms through which HIV accelerates the progression of liver disease due to HCV are not completely understood.[20] In this sense, the effects of HIV infection on the immune response to HCV may play a pivotal role.[21] Moreover, HIV coinfection may lead to increased hepatocyte apoptosis,[22] which could worsen liver fibrosis.[23] Finally, some hepatotoxic antiretroviral drugs commonly used in HCV/HIV-coinfected patients, such as nevirapine, may enhance the progression of liver fibrosis.[24] Probably as a consequence of these deleterious effects, HIV carriers showed signs of more severe liver disease, through a poorer Child-Turcotte-Pugh score and MELD score, at the time of the first hepatic decompensation. That notwithstanding, the fact that HIV infection is a mortality predictor independent of the aforementioned scores suggests that the negative impact HIV exerts on the outcome of HCV-related liver disease before the development of decompensated cirrhosis[4][5] is also maintained in ESLD.
In this study, HCC was included as a form of hepatic decompensation because it can be considered a complication of cirrhosis. In other studies in which the survival of decompensated cirrhosis was assessed, HCC was not included as a type of decompensation.[25][26] This fact might have led to a reduction in the survival of the patients included in this survey. However, the survival of HIV-uninfected patients in this study at 5 years was 44%, a figure close to the 51% reported by Planas et al.[25] and the 50% found in the EUROHEP cohort.[26] Moreover, the presence of HCC did not predict survival - although if HCC had affected mortality, the differences between HIV-infected and HIV-uninfected patients would have been even larger after adjusting for HCC, because it was much more frequent in the latter group.
The independent risk factors for death - other than HIV infection - found in this study are not surprising. Thus Child-Turcotte-Pugh score, MELD score, and the number of hepatic complications are markers of liver disease severity. These parameters, as well as age, have been found to be prognostic variables in decompensated HCV-related cirrhosis in patients without HIV infection.[25-27] HDV coinfection is associated with a poorer prognosis in decompensated cirrhosis in patients infected with hepatitis B virus.[28] Likewise, hepatitis B virus infection has been proven to be associated with a higher risk of ESLD in HCV-infected patients in other studies.[4][5] Because of this, an additive liver damage caused by these three viruses is conceivable and would explain the effect of HDV on survival found in this study. HE as the first decompensation has also been found previously to have a strong impact on the survival of patients with cirrhosis,[29][30] including those harboring HCV-related cirrhosis.[25] In our study, HE was associated with a poorer outcome independently of the number of simultaneous events patients presented. Moreover, only 4 of 115 (4%) patients with HE as the first hepatic decompensation showed PHGB simultaneously (data not shown). These findings suggest that HE itself determines a poor prognosis and that this effect is not a consequence of a sum of events.
The lack of an association between alcohol drinking during follow-up and mortality is relatively surprising, because alcohol seems to enhance the liver damage caused by HCV.[31] Nevertheless, findings similar to those observed in this study have been reported by other authors.[25] Moreover, many surveys have failed to find any association between alcohol consumption and liver fibrosis.[24][32-34] In the present study, as in many others, the alcohol consumption calculation was based on the clinical interview. This procedure of estimating daily alcohol intake is very unreliable in alcohol drinkers.[35] Because of this, underreporting or misclassification of alcohol consumption is likely.
Another remarkable finding in this study was that the frequency of HE as both the first decompensation and the cause of death was higher in HIV-coinfected patients. HE is a serious complication of liver disease and is usually included in prognosis scores as an ominous parameter. Because of this, that finding is consistent with more severe liver disease in HIV-coinfected patients. However, most HIV-infected patients were drug users. Because of this, they could have consumed depressors of the central nervous system more commonly than HCV-monoinfected individuals, which may have hastened HE episodes. The data of this study indicate that the incidence of HE may be increased in HIV-coinfected patients with cirrhosis due to HCV. Likewise, the fact that HE is more frequent as the first decompensation suggests that it may occur earlier in HIV patients. Studies aimed at acquiring insight into these topics - and, if confirmed, into the underlying mechanisms - are needed. By contrast, the frequency of HCC was lower among HIV-coinfected individuals. This finding conflicts with that of a case-control study in which HCC emerged after a shorter period of HCV infection in subjects coinfected with HIV.[36] However, HIV coinfection may shorten the survival of HCV-related liver disease carriers so much so that patients do not live long enough to develop HCC, a late complication of HCV infection. Thus the median length of HCV infection in 24 HIV-uninfected patients with HCC as the first decompensation in whom this time could be estimated was 22 years. Only 12 of 158 (8%) HIV-infected subjects survived this time after the estimated date of HCV infection (data not shown). The results of this study regarding the frequency of HCC in HIV/HCV-coinfected patients are in agreement with those observed in other cohorts in which HCC was not observed[7] or the incidence was decreased.[8]
The protocols currently used for including HIV-infected patients with ESLD in liver transplantation programs consider that the liver disease criteria for transplantation should be the same as those for the general population.[11][37] However, these criteria are based on survival studies performed in patients without HIV infection. At least in the case of HCV-related ESLD, this recommendation may be inappropriate, given the effect that HIV coinfection exerts on the survival of this disease. In a small series of Spanish HIV-infected patients who underwent liver transplantation, the average time spent on a waiting list was 3 months (range: 1-12).[38] However, because liver transplantation is regarded as a suitable procedure for HIV-infected patients, wait times are expected to increase. In summary, according to the results reported herein, many HIV-infected patients with HCV-related ESLD would not survive long enough to obtain an available organ. This can be particularly true if more risk factors for poor outcome are present, such as Child-Turcotte-Pugh class B or C, high MELD score, HE, several simultaneous complications at onset of decompensated disease, or HDV infection. Because of this, specific studies on the predictors of survival in HIV-infected patients with ESLD are urgently needed. These studies will allow clinicians to establish the best time to put these patients forward for a liver transplantation program. Likewise, strategies to make liver transplantation available as soon as possible after a patient is allocated to this procedure will be required.
Results
Characteristics of the Study Population.
One thousand two hundred seventeen HCV-infected patients with decompensated cirrhosis and fulfilling the inclusion criteria attended the participating hospitals during the study period; 180 of these (15%) were seropositive for HIV. The most relevant demographic, virological, and clinical characteristics of HIV-coinfected and HIV-uninfected patients are shown in Table 1. In 218 (21%) HCV-monoinfected patients and in 41 (23%) HIV-coinfected subjects (P = .59), the diagnosis of cirrhosis had been made via biopsy. HIV infected were more likely to be:
IDUs; higher daily alcohol intake; HBsAg+; HDV+; Child Turcotte Pugh Class C; MELD higher score (average: 16 vs 12).
The median follow-up of the population studied was 13 months (interquartile range: 4-33). Seventeen (9%) HIV-infected patients and 891 (86%) individuals without HIV infection were followed at Hepatology units. The remaining subjects received care at Internal Medicine or Infectious Diseases units. Eighty-seven (7%) patients, all of them HIV-seronegative, underwent a liver transplant. Seventy-seven patients (43%) with HIV infection were prescribed highly active antiretroviral therapy. Fifty-two (29%) HIV-infected patients were in Centers for Disease Control and Prevention class A, 43 (24%) were in class B, and 85 (47%) were in class C. The median CD4+ cell count of HIV-coinfected patients at baseline was 191 cells/L (interquartile range: 75-342). Plasma HIV RNA load determinations were available in 163 HIV-infected patients at the time of the first hepatic decompensation. Twenty-four (15%) patients showed concentrations below 200 copies/mL. The median value in individuals with detectable levels was 4.16 log copies/mL (interquartile range: 2.84-4.92).
The median follow-up of the population studied was 13 months (interquartile range: 4-33). Seventeen (9%) HIV-infected patients and 891 (86%) individuals without HIV infection were followed at Hepatology units. The remaining subjects received care at Internal Medicine or Infectious Diseases units. Eighty-seven (7%) patients, all of them HIV-seronegative, underwent a liver transplant. Seventy-seven patients (43%) with HIV infection were prescribed highly active antiretroviral therapy. Fifty-two (29%) HIV-infected patients were in Centers for Disease Control and Prevention class A, 43 (24%) were in class B, and 85 (47%) were in class C. The median CD4+ cell count of HIV-coinfected patients at baseline was 191 cells/L (interquartile range: 75-342). Plasma HIV RNA load determinations were available in 163 HIV-infected patients at the time of the first hepatic decompensation. Twenty-four (15%) patients showed concentrations below 200 copies/mL. The median value in individuals with detectable levels was 4.16 log copies/mL (interquartile range: 2.84-4.92).
HIV infected were;
Younger 38 vs 68 yrs
More likely to be IDUs
More likely to have greater daily alcohol intake
Causes of Death.
In 225 (18%) individuals, 13 (7%) HIV-infected and 212 (20%) uninfected, the life status could not be determined at the close of the study. The vital status and, when appropriate, the cause of death, were obtained through the regional death registry in 14 (8%) HIV-infected and 73 (7%) HCV-monoinfected patients and via telephone contact in 3 (2%) HIV-coinfected and 31 (3%) HIV-uninfected subjects. In the remaining cases, these data were obtained from clinical records or hospital databases. Autopsies were available in 8 (4%) HIV-coinfected and 34 (3%) monoinfected individuals.
Three hundred eighty-six (37%) patients without HIV infection and 100 (56%) HIV-infected individuals died during the follow-up. Liver disease was the cause of death in 332 (86%) monoinfected individuals and in 81 (81%) coinfected patients; AIDS accounted for 5% of deaths in the latter group. The frequencies of the events that led to death due to liver disease were different in patients with and without HIV infection. Specifically, HE and spontaneous bacterial peritonitis were more frequent among HIV-coinfected individuals, whereas HCC accounted for a greater percentage of cases among monoinfected patients.
The median survival was shorter in patients harboring baseline plasma levels of alpha fetoprotein greater than or equal to 20 ng/mL (30 vs. 63 months; P < .001). However, this parameter did not remain associated with survival after adjusting for the presence of HCC throughout the follow-up. Thus the median survival of 553 patients without HCC and a baseline alpha fetoprotein plasma level of below 20 ng/mL was more than 72 months, whereas this value was 60 months in 169 patients without HCC and alpha fetoprotein plasma levels of 20 ng/mL or more (P = .16). In 215 patients who developed HCC throughout the entire follow-up, the corresponding median survivals were 37 months and 11 months, respectively (P = .003).
The unit that provided care to the patient, the route of HCV infection, prior interferon therapy (either alone or with ribavirin), and the use of intravenous drugs during the follow-up also predicted survival in the univariate analysis (see Table 4). Conversely, gender, daily alcohol intake, HBV coinfection, HCV genotype, plasma HCV RNA load, the procedure of cirrhosis diagnosis, and the hospital where the patient was recruited and followed were not associated with survival.
In the multivariate analysis, HIV infection, a poorer Child-Turcotte-Pugh class, MELD score, more than one event at the first hepatic decompensation, older age, and seropositivity for anti-HDV at baseline - as well as the presence of HE as the first decompensation of cirrhosis - independently predicted survival (see Table 4).
The impact of the length of HCV infection on survival was assessed in the subgroup of 338 patients in whom this time could be estimated. The median survival in patients with a duration of HCV infection longer than 18 years was 44 months, whereas it was 30 months in those who had been infected for 18 years or less (P = .01). However, this parameter did not remain as an independent predictor of survival in a Cox regression analysis limited to this subgroup of patients (P = .41).
An analysis restricted to the subgroup of patients in whom survival data were collected from clinical records or databases yielded the same results as that performed in the entire population.
INTRODUCTION
It is well known that human immunodeficiency virus (HIV) coinfection adversely affects the course of hepatitis C virus (HCV)-induced liver damage. The progression of liver fibrosis is faster in coinfected patients than in HCV-monoinfected individuals.[1-3] Moreover, HIV infection accelerates the emergence of decompensated HCV-related cirrhosis.[4][5] Likewise, mortality due to liver disease is increased in patients with chronic hepatitis C and HIV coinfection.[6] The mortality attributable to hepatocellular carcinoma (HCC) has been shown to be increased in this setting,[6] but controversial results have been reported concerning the incidence of HCC in the coinfected population.[7][8] Few data are available regarding the impact of HIV coinfection on the survival of patients harboring HCV-related end-stage liver disease (ESLD). According to the above-mentioned facts, a reduced survival could be also expected at this stage. Nevertheless, the effect of HIV coinfection may be minimized in advanced liver disease due to other factors that could have a stronger impact on the survival of patients with cirrhosis; these include age, alcohol consumption, appropriate medical care, hepatitis B virus (HBV) coinfection, and a poor Child-Turcotte-Pugh score.
As survival of HIV-infected persons has improved due to widespread use of highly active antiretroviral therapy, mortality due to HCV-related liver disease has increased in this setting. Consequently, cirrhosis caused by HCV has become a major cause of death in some areas.[9][10] These facts have led some experts to consider liver transplantation as a suitable therapeutic tool in coinfected patients with HCV-related ESLD.[11][12] The results recently reported in a case series dealing with the outcome of HIV-infected liver transplant recipients suggest that survival does not differ from that of HIV-negative patients.[13] This finding supports the indication of liver transplantation in HIV/HCV-coinfected individuals with ESLD. If HIV-infected patients are to be considered candidates for liver transplantation, it is required to know whether HIV coinfection reduces the survival of coinfected patients with HCV-related ESLD. This may help us to establish priorities among transplantation candidates rationally.
The purpose of this study was to compare the survival of HIV-infected and HIV-uninfected patients with HCV-related cirrhosis after the first hepatic decompensation.
Patients and Methods
Study Design and Data Collection.
Data on survival from a cohort of patients with decompensated HCV-related cirrhosis who attended eight hospitals in Andalusia (in southern Spain) and who were followed thereafter were analyzed retrospectively in this survey. Eligible patients were identified by using the database of the clinical documentation services of each hospital. The codes of the International Classification for Diseases (ICD-9) for cirrhosis, each type of hepatic decompensation, viral hepatitis, and liver primary carcinoma were used as search terms. A common database was created to collect the data. The study was designed to detect a difference of 18 months in the median survival time (30 vs. 48 months), with a type I and type II error rate of 0.05 and 0.1, respectively. The computed accrual time of patient recruitment was 6 years, and the minimum follow-up was 1 day. A ratio of control to cases of 6 was considered. The minimum sample size thus calculated was 1,079 patients. The program PS Power and Sample Size Calculations, version 2.1.30 (http://www.mc.vanderbilt.edu/prevmed/ps/index.htm), was used for this calculation.
Patients and Follow-Up.
In this study, patients with decompensated cirrhosis who was seen at each participant hospital were analyzed from January 1997 to December 2002, provided that they fulfilled the following inclusion criteria: (1) older than 18 years of age; (2) positive serum anti-HCV; (3) detectable plasma HCV-RNA; (4) no evidence of metabolic or autoimmune liver disease according to clinical history, appropriate laboratory tests, and - when available - histological examination; and (5) no previous decompensation of liver disease. All eligible patients who were consecutively seen at the participant hospitals in the study period, regardless of HIV serostatus, were included in the analysis.
During the follow-up, all patients received similar care. The guidelines of the Spanish Association for the Study of the Liver (www.aeeh.org) were followed by both hepatologists and nonhepatologists. Patients were assessed at least every 6 months for symptoms and signs of hepatic decompensation. Ultrasound abdominal examinations for HCC screening were performed at least every 6 months. Portal hypertensive gastrointestinal bleeding (PHGB),[14] ascites, hepatorenal syndrome,[15] and spontaneous bacterial peritonitis[16] were managed according to reported consensus. Namely, primary prophylaxis for PHGB was performed using beta-blockers (nadolol or propanolol). After PHGB, beta-blockers plus nitrates were used. Norfloxacin was administered in patients with spontaneous bacterial peritonitis for life and in those with an episode of PHGB for 1 week. No treatment other than lactitol or lactulose was prescribed after bouts of hepatic encephalopathy (HE). Liver transplantation was considered in HIV-uninfected patients with Child-Turcotte-Pugh class B or C and no contraindications. Likewise, transplantation was regarded as a therapeutic option in patients with HCC and unique tumors or three or fewer nodules with a diameter of less than 3 cm and no extension data.
All HIV-infected individuals who survived the first hepatic decompensation episode and who were candidates for HIV therapy were prescribed highly active antiretroviral therapy throughout the follow-up. Combinations included nucleoside analogues, protease inhibitors, or nonnucleoside retrotranscriptase inhibitors, as they became available. Drug combinations were made according to the recommendations of international guidelines applicable at the time of prescription.
In patients who acquired HCV through the sharing of needles or via blood transfusion, the duration of HCV infection was estimated. This value was calculated as the time elapsed from either the date of the first blood transfusion or the first year of needle-sharing to the date of the first liver decompensation.
Diagnostic Criteria.
Cirrhosis was diagnosed via histological examination or by the combination of clinical, biochemical, and ultrasound imaging data agreeing with such a diagnosis. Cirrhosis cases that presented with ascites, PHGB, HE, nonobstructive jaundice, spontaneous bacterial peritonitis, or HCC were classified as decompensated. The event that led the patient to the hospital was considered as the type of the first hepatic decompensation in the analysis. When more than one complication was present simultaneously, what dominated the clinical picture in the opinion of the attending physician was considered as the first hepatic decompensation.
Ascites was always proven by ultrasound examination or paracentesis. PHGB was diagnosed according to the Baveno criteria.[14] HE was defined as an episode of mental confusion related by the patient or their relatives, or detection of disorientation by a physician in the absence of any other likely nonhepatic causes. Patients were classified as having nonobstructive jaundice if the plasma bilirubin level was equal or greater than 5 mg/dL and the ultrasound examination showed normal bile ducts. Spontaneous bacterial peritonitis was diagnosed when the polymorphonuclear count of the ascitic fluid was greater than 250 cells/L or bacteria grew in the ascitic fluid culture, and no clinical or image data of secondary peritonitis were present. The diagnosis of HCC was made according to the criteria proposed by the panel of experts of the European Association for the Study of the Liver.[17] Hepatorenal syndrome was defined as proposed by Ginˇs and colleagues.[18]
The model for end-stage liver disease (MELD) score was calculated using the following formula: 3.8Lnbilirubin + 9.6Lncreatinine + 11.2LnINR + 6.4.[19]
Life Status and Cause of Death Determinations.
To establish the vital status of the patients and the cause of death, the clinical records and the database of each participant hospital were used. Persons lost to follow-up or their next of kin were contacted via telephone whenever possible. The regional death registry was consulted to obtain the date and cause of death when a patient was lost to follow-up.
Death was classified as liver-related if it was a consequence of HCC, PHGB, hepatorenal syndrome, HE, or spontaneous bacterial peritonitis or whenever liver failure had a key role in death. AIDS was considered the cause of death when there was a life-threatening disease defining the clinical stage C of the 1993 Centers for Disease Control and Prevention classification. Otherwise, the cause of death was categorized as other.
Laboratory Procedures.
Serum HCV antibodies were identified via enzyme immunoassay (EIA; Ortho Diagnostic System, Raritan, NJ). HCV-RNA was detected either via qualitative polymerase chain reaction (Amplicor HCV or Cobas Amplicor HCV; Hoffmann-La Roche, Basel, Switzerland) or quantitative polymerase chain reaction (Amplicor HCV Monitor or Cobas Amplicor HCV Monitor; Hoffmann-La Roche). HCV genotype was determined via line-probe assay (INNOLiPA HCV; Innogenetics, Ghent, Belgium). HIV antibodies were tested via enzyme immunoassay (Genscreen HIV-1/2; Pasteur Diagnostics, Paris, France) and confirmed via Western blot analysis (New LAV Blot I; Pasteur Diagnostics). Plasma HIV RNA was measured via polymerase chain reaction assay (Amplicor HIV Monitor test; Hoffmann-La Roche). Serum hepatitis B surface antigen was detected via enzyme immunoassay (HBsAg Abbott Axsyn System; Abbott Cient’fica S.A., Madrid, Spain). Hepatitis D virus (HDV) antibodies were also tested via enzyme immunoassay (Delta Bioelisa HDV; Biokit, Barcelona, Spain).
Statistical Methods.
Continuous variables are expressed as the median (Q1-Q3) and survival times as median. When the median survival was not reached, a value of more than 72 months is given, because 72 months was the maximum possible follow-up in this study. Categorical variables are presented as numbers (percentage). Survival estimates at different time points are expressed as the cumulative proportion of survivors at the end of the period. Comparisons between categorical variables were made using the 2 test or the Fisher test. For continuous variables, the Student t test was used if homocedasticity and normal distributions were proven. The Mann-Whitney U test was used otherwise.
For survival analysis, data were censored at the date of death or when the last information on life status was obtained and no later than December 31, 2002. For patients who underwent liver transplantation, the date of censoring was that of the surgical procedure.
The survival time was calculated from the date of the first decompensation to death or censoring date. The univariate analysis of survival was performed using the Kaplan-Meier method. Curves were compared using the log rank test. Continuous variables were dichotomized taking the median as the cutoff value, except for alpha fetoprotein, whose cutoff value was 20 ng/mL. Estimates of survival at different times of the follow-up were obtained using life tables. Independent prognostic factors for survival were identified by a stepwise forward Cox regression model, where covariates were entered if a difference with P < .1 was detected in the univariate analysis. MELD score was treated as a continuous variable in the Cox regression.
The statistical analysis was performed using the SPSS statistical software package, version 11 (SPSS, Chicago, IL).
Ethical Aspects.
The study was designed and performed according to the Helsinki Declaration and was approved by the ethics committee of each participating hospital.
|
|
| |
|
|
|
