Factors of Prognostic Importance in Primary Biliary Cirrhosis
Scand J Gastroenterol Downloaded from informahealthcare.com by University of British Columbia on 10/29/14 For personal use only.
A. RYDNING, E. SCHRUMPF, M. ABDELNOOR, K. ELGJO & E. JENSSEN Medical Dept. A and Dept. of Pathology, Rikshospitalet, The National Hospital, and Institute of Medical Statistics, Ullevll Hospital, Oslo, Norway Rydning A, Schrumpf E, Abdelnoor M, Elgjo K, Jenssen E. Factors of prognostic importance in primary biliary cirrhosis. Scand J Gastroenterol 1990, 25, 119-126 To determine survival and the risk factors of death in primary biliary cirrhosis, data from 52 symptomatic and 13 asymptomaticpatients were analyzed. The mean followup time was 6.3 years (range, 0.4-23 years). The average length of survival was 18 years for the symptomaticand 8.4 years for the asymptomaticpatients. By a univariate analysis, ascites, presence of esophageal varices, gastrointestinal bleeding, jaundice, hepatomegaly and the logarithms of albumin and bilirubin were all associated with a poor prognosis. A multivariate analysis of the clinical features showed that the presence of bleeding from esophageal varices and the logarithm of bilirubin were the only predictors for poor prognosis. The survival of the symptomaticpatients is longer than reported previously, while the life expectancy for the asymptomatic patients seems no better than for the symptomatic group. Key words: Clinical features; histology; primary biliary cirrhosis; prognostic factors;
survival Andreas Rydning, M . D . , Medical Dept. A , The National Hospital, 0027 Oslo I , Norway
Primary biliary cirrhosis (PBC) is a chronic cholestatic liver disease of unknown etiology. Without hepatic transplantation the terminal outcome is death in liver failure or bleeding esophageal varices. The prognosis for liver diseases has until recently been assessed largely on the basis of clinical criteria and the use of Child’s classification, often with Pugh’s modification (1). Although serum bilirubin has been shown to be a good marker of progression of disease ( 2 4 ) , other factors have also been shown to be of importance (4,7-9). Today the diagnosis of PBC is often made in asymptomatic patients, probably because of the more elaborate use of liver function tests on routine examinations (10). Symptoms and progressive disease eventually develop in some of these patients. The aim of the study has been to look at survival and determine factors of prognostic importance in patients with symptomatic and asymptomatic PBC. Since liver transplantation is a new thera-
peutic modality in PBC, is is important to determine the natural course of the disease and determine risk factors of mortality, to optimize the timing of transplantation. MATERIALS A N D M E T H O D S During the years 1970-1986, 65 patients with clinical, biochemical, and histologic evidence of PBC were admitted to Medical Dept. A , The National Hospital, Oslo, Norway. A t the time of admission 52 patients (80%) had symptoms of liver disease, whereas 13 (20%) patients were asymptomatic. Patients were regarded and classified as asymptomatic if no symptoms could be attributed to their liver disease. Signs and symptoms more specifically related to PBC include pruritus, icterus, hepatomegaly and/or splenomegaly, and ascites. The non-specific symptoms are malaise o r fatigue, abdominal pain, hyperpigmentation, weight loss, and xanthomas (2-
Scand J Gastroenterol Downloaded from informahealthcare.com by University of British Columbia on 10/29/14 For personal use only.
120
A . Rydning ct ul.
4.6). Other symptoms not related to the liver disease were not included in estimating the start point of the disease. Most of the asymptomatic patients were identified on the basis of a raised serum level of alkaline phosphatase and/or alanine aminotransferase detected during a routine physical examination. Coexisting autoimmune diseases were recorded. In the same period 21 patients with cholestatic !iver disease of uncertain origin were admitted. 'These patients were classified as follows: secondary biliary cirrhosis, cholestatic liver disease, or chronic active hepatitis/PBC?
Nistology Liver biops,y sections were examined by one of [he investigators (K. Elgjo) at the time of initial cv;iluation and re-examined again now. The diagnosis was established and classified in accordance with Scheuer (1 1).
('linical and laboratory data Data on symptoms, physical findings, and laboratory data were obtained from hospital records at the time of diagnosis. The laboratory data analyzed were alkaline phosphatase, alanine aminotransferasc, bilirubin, albumin, prothromhin time. cholesterol, IgM, antinuclear antibody (ANA), smooth-muscle antibody (SMA), and aritimitochondri;rI antibody (AMA). All patients had patent biliary ducts demonstrated by intravenous cholangiography, endoscopic retrograde chol;rngiography, or operative cholangiography. Survioml Information on survival o r death of any patient was obtained from The National Center for Health Statistics. The termination of observation was 1 October 1987. The mean follow-up time was 7 years (range, (1.7-23 years) for the group o f symptomatic patients and 5.7 years (range, 0.4-1 1 . 1 years) for the asymptomatic patients. ,S(afi.sticulanaJysis A preliminary analysis with a univariate method was performed for each postulated progiiostic factor in the symptomatic group of patients, using survival curves. Continuous variables were
tested with the Mann-Whitney or Students' t test. Discrete variables were tested with the contingency table. Comparison of univariate survival curves was made with Breslow (12) and Mantel-Cox test statistics (13). To estimate risk factors for total mortality, the data were analyzed with the Cox regression model (14). The advantages of this model are linkage of survival function to prognostic variables and the possibility of quantifying the effect of prognostic covariates on survival. The essential check of the proportional effect of significant covariate was met by plotting the logarithm of the integrated hazard. The variables estimated by single-factor analysis were subjected to multifactorial analysis by manual backward elimination procedures. Multifactorial procedures were preceded by estimation of correlations between variables. If some variables were significantly correlated, their predictive value was assumed to be similar. The prognostic factors identified as independent were tested for interactions. A BMDP statistical program and an IBM 4341 LO1 computer were used. RESULTS The ratio between male and female patients was similar for both the symptomatic (21 % male, 7Y% female) and the asymptomatic group (15% male, 85% female), and there were no significant differences in age between the two groups at the time of diagnosis (53.1 and 51.3 years, respectively). Our patients were followed up for a mean of 6.3 years, a total of 6.3 X 65 = 409 patient-years. The observed frequency of signs and symptoms is listed in Table I . Symptoms Pruritus was the most frequent initial symptom among the symptomatic patients, experienced in nearly SO% of the cases (Table I). Malaise and abdominal pain were the second and third most frequently registered presenting complaints. Several patients had more than one initial symptom. Five (38%) of the asymptomatic patients developed symptoms during the time of observation (mean, 5 years; range, 2-8 years). Two patients experienced pruritus; one, abdominal pain plus
Prognostic Factors in PBC
121
Scand J Gastroenterol Downloaded from informahealthcare.com by University of British Columbia on 10/29/14 For personal use only.
Table I. Frequency of signs and symptoms in 65 patients with PBC
Symptoms/signs
Symptomatic, % ( n = 52)
Asymptomatic, % ( n = 13)
P
Pruritus Malaise/fatigue Abdominal pain Weight loss Anorexia Nausea Hepatomegaly GI hemorrhage Jaundice Splenomegaly Hyperpigmentation Ascites Xanthomas
49 47 45 37 21 19 53 25 25 17 9 8 4
31
NS
15
NS
8
NS
Table 11. Associated autoimmune disorder in 52 symptomatic and 13 asymptomatic patients with PBC n Thyroid disease Keratoconjunctivitis sicca Scleroderma Disseminated lupus erythematosus Raynaud’s phenomenon Antibody towards salivary gland Congestive cardiomyopathy Rheumatoid arthritis
96
14 5 3 1.5 1.5 1.5 1.5 1.5
malaise; one, ascites, icterus, and dyspnea; and the last patient developed variceal bleeding and coma.
Signs A m o n g the physical signs registered at the time of diagnosis, hepatomegaly was by far the most
frequent sign seen both among the symptomatic (53%) and the asymptomatic (31%) patients (Table I). A t diagnosis of PBC gastrointestinal hemorrhage was seen in 25% of symptomatic and in none of the asymptomatic patients, while jaundice was equally present in the two groups.
Autoimmune phenomena Autoimmune phenomena were present in 17 patients (12 symptomatic and 5 asymptomatic) (Table 11). Three patients had more than one autoimmune phenomenon. Laboratory values (Table I l l ) Mean level of alkaline phosphatase was 675.1 U/1 in the symptomatic and 830.4 U/1 in the asymptomatic group. T h e same tendency with higher levels in asymptomatic patients was seen with regard to alanine aminotransferase ( A L A T ) . Mean serum bilirubin was 29.9 pmol/l in the
Table 111. Laboratory data in 65 patients with PBC Symptomatic
Alkaline phosphatase (55-195 U/1) Alanine aminotransferase (9-40 U/I) Bilirubin (3-26 pmol/l) Albumin (35-50 g/l) Prothrombin time (65-150%) Cholesterol (2.S6.5 mmol/l) IgM (0.5-2.8 g/l)
Asymptomatic
n
Mean
SD
n
Mean
SD
52 52 52 51 51 39 49
676 75 30 34 101 7.2 6.4
539 58 34 8.7 33 3.2 5.1
13 13 13 13 10 11 12
830 96 17 30 118 7.5 7.0
455 80 12 17 34 3.7 3.8
122
A. Kydning et ul.
Tahlc IV. Percentage of positive autoantibodies in 65 patients with PBC I _
Autoantibody
Symptomatic ( n = 52)
Asymptomatic (n = 13)
96 42 2s 9
92 50 46 15
Scand J Gastroenterol Downloaded from informahealthcare.com by University of British Columbia on 10/29/14 For personal use only.
Mitochondria1 antibody l i t e r 2 512 Smooth-muscle antibody Antinuclear antibody
1';rhlc V. Histologic grading o f the first representative liver biopsy in symptomatic and asymptomatic PBC Symptomatic ( n = 52)
Iflstoioglc stage
Asymptomatic ( n = 13)
P NS NS NS
NS
A M A were positive in nearly all o f the patients regardless of symptoms (Table IV). The numbers of patients with SMA and A N A were similar in the two groups of patients.
I I1
4 35 1
symptomatic and 16.6 pinol/l in the asymptomatic group. None of the differences above reached statistical significance. Nor were there any statistical differences between the two groups in levels of albumin, prothrombin time, total cholesterol, and total IgM. Equal numbers of patients had a raised IgM level (IgM > 2.8 g/l) in the symptomatic and the asymptomatic group.
Operations Twelve symptomatic patients were operated on with portosystemic shunts during the follow-up period. Ten patients received a portacaval shunt and two patients received a splenorenal shunt. Four symptomatic patients received a liver trnnsplant. They were followed up until the time when they received a new liver. Histology (Table V ) Among the 52 symptomatic patients there was 1 patient from whom we did not obtain a biopsy, 1patient with normal histology, and 35 with histo-
% Survival
30 -
-
20 10
m.
-
I
7-m
I
2
i
4
I
6
I
8
1
I
i
0
15
m m
Nomslpapul Symptomatic A$ymplomatii
I
I
2o
Years
25
Fig. I . Survival curves for the 52 symptomatic and the 13 asymptomatic patients with PBC.
Prognostic Factors in PBC
123
factors. The presence of ascites resulted in a lower 5-year survival (25%) than in patients without this sign (85%) (Table VI). Patients with varices had a pool survival of 44% versus 88% for patients without. Jaundice gave a pool survival of 75% versus 88%. The presence of gastrointestinal bleeding caused a lower survival rate, 54%, versus 90% for those without this sign. Patients without abdominal pain seemed to have a lower survival Survival Survival curves were drawn for the symp- rate than patients who did experience pain (72% tomatic and the asymptomatic group as shown in versus 92%). Hepatomegaly resulted in poorer Fig. 1. There were 17 (33%) out of 52 PBC- survival, 74% versus 88% in those with a normalrelated deaths in the symptomatic group. Survival sized liver. The other dichotomized variablestimes were calculated from the time of onset splenomegaly , presence of xanthomas, presence of symptoms until death. The observed mean of anorexia, nausea, hyperpigmentation, pressurvival time was 18 years for the symptomatic ence of itching, and loss of weight-did not influgroup. There were five deaths not related to liver ence survival prognosis. The liver histologic data did not influence survival either. disease. A two-sample t test o r the Mann-Whitney test In the asymptomatic group four (31%) deaths were related to the liver disease. O n e additional was carried out, comparing the continuous varideath was not related to liver disease. Survival ables, depending on whether the distribution of times for this group of patients were calculated the variables was normal o r skewed. Age at diagfrom the time of diagnosis, which usually took nosis came out as a significant factor (Table VII). place at their first hospital admission. The After studying the distribution of these factors, observed mean survival for this group of patients we performed a logarithmic transformation, since most of the values had a high degree of skewness. was 8.4 years (Fig. 1). The logarithms of albumin and bilirubin were both significantly different among patients who Univariate analysis Patients who survived were compared with died and patients who are still alive (Table VII), patients who died, with regard to the following whereas the logarithms of alkaline phospatase,
Scand J Gastroenterol Downloaded from informahealthcare.com by University of British Columbia on 10/29/14 For personal use only.
logic changes in accordance with Scheuer grades 111 and IV. Among the 13 asymptomatic patients there were 9 with Scheuer grade I11 and IV changes and only 3 with grade I and I1 changes. In one patient the histologic material was insufficient for evaluation.
Table VI. Clinical factors of prognostic importance for survival (univariate analysis) in 52 symptomatic PBC patients Sign/symptom (variable) Ascites Varices
GI bleeding Jaundice Abdominal pain Hepatomegaly
Cumulative 5-year survival -+ SEM
Chi-square Breslow (PI
Chi-square Mantel-Cox (P)
15.4 (0.0001)
17.6 (0.0001)
11.6 (0.0001)
11.8 (0.0001)
43
25 2 0.22 85 t 0.05 44 5 0.17 88 2 0.05
39
54 2 0.14 90 -t 0.05
9.7 (0.002)
9.0 (0.003)
25
75 2 0.15 88 2 0.07
7.4 (0.007)
11.4 (0.0007)
92 2 0.06 71 2 0.09 74 -t 0.09 88 2 0.07
5.8 (0.015)
4.7 (0.03)
4.1 (0.04)
4.4 (0.03)
Present, yes/no
4 48 9 13 8 24 28 27 25
I'ithle VII. (.'linical and laboratory values of prognostic importance (continuous variables) in 52 symptomatic puticnts
Alive
v.(11."i'ible (log)
n
Mean
SD
n
Mean
SD
I'
Agc at diagnosis
3.5 34 35
1.72 1 .so 1.20
0.0')
17 17 17
1.77 I .4X 1 .55
0.09
0.009 0.002 0.005
Albumin l3ilirul)in Scand J Gastroenterol Downloaded from informahealthcare.com by University of British Columbia on 10/29/14 For personal use only.
Dead
0.26
0.34
0.08 0.29
i';ihle VI11. Results of thc inultivariatc Cox modcl analysis: risk factors of mortality ~~
~
Variables Presence of bleeding 1-og hilirubin
Level
73
SE(t%
Y es/no
1.6812 2.0274
0.5577 0.7 101
~
.41..4'1~.prothrombin time, and cholesterol were not. Another univariate analysis was performed rising t h c C o x inodel and the logarithms of the continuous variables. Among the continuous variables. the logiarithm of bilirubin was highly \igi.rificant (p = O.Ol), whereas the logarithm of AIAT did n o t reach statistical significance ( p = 0.09).
Mitltiouriute irnidy.si.s After ;I tinivariate analysis of the different postulated prognostic factors had been completed, these I ? covariates were pooled together in a multivariate an;ilysis by the proportional hazard iiiodel. A multivariate backward elimination proivtitire was u:;ecl. After different steps estimating thc Iikclihood of the model, seven steps were run until the most stable model was established. The
&S E (
p)
3.0 145
2.8551
presence of gastrointestinal bleeding and the logarithm of bilirubin proved to be independcrit predictors of mortality. The model for the relative risk o f mortality (RR) (Table VIII) in patients with PBC was derived as follows:
RR
= (,I.h8I?(hkcdlng
(I.:?)
+ 2,(i771~,1,ihlllrllhln
DISCUSSION The patients included in this retrospective study had been admitted from all of Norway. although most of them were from the southeast part of the country.
Prognostic I'rcsencc of G I bleeding
Case A
Case B
Case C
Case D
NO
YES
NO
YES
31.3
31.3
0.65
3.52
('oncentration of bilirubin (pmol/l)
100
100
liclativc risk 01 mortality of I'BC
31) i l
By means of this model, the relative risk5 for four hypothetical patients have heen forecast ('Table IX).
I ;ibk IM. Relative risk of mortality Eor lour patients with different subsets of prognostic variables
variable
~
27.5
147
Scand J Gastroenterol Downloaded from informahealthcare.com by University of British Columbia on 10/29/14 For personal use only.
Prognostic Factors in PBC
Nyberg & Loof (8) found that 70% of their 80 patients were asymptomatic at initial presentation, in contrast to the report from Sherlock & Scheuer (9) in 1973, which described the initial presentation of PBC as asymptomatic in 11% of the cases. In our study 20% of patients were initially asymptomatic. The difference in ratio between asymptomatic and symptomatic patients in these studies is most likely related to the period of inclusion. The Swedish study is a follow-up from the years 1979-1986, when more elaborate laboratory data and health investigations were performed, while our first inclusion was in 1970. Others ( 4 , 5 , 9 ) have confirmed that there has been a shift from symptomatic to asymptomatic PBC during the past decade, most likely due to the increased use of laboratory profiles. Several authors have suggested a better prognosis for those patients who are asymptomatic at the time of diagnosis. Roll et al. (4) and Beswick Kr Boyer (5) found a survival rate for asymptomatic patients equal to that of an age- and sexmatched cohort of the general population. Nyberg & Loof (8) report a 50% survival of 16 years in asymptomatic PBC, the reduced survival rate appearing more marked after 12 years of observation. Balasubramarriam et al. (15) published similar poor survival figures for asymptomatic patients, with a drastic increase in mortality appearing after 5 years. In our limited number of asymptomatic patients the survival rate was no better than for the symptomatic patients. Survival may be influenced by the difference in the starting point for registration in the two groups, since the asymptomatic group was registered from the time of diagnosis, and the symptomatic group was registered from the time the symptoms started. Differences in survival times reported between different authors may be due to different criteria chosen for symptomatic/asymptomatic PBC. Nyberg & Loof (8) d o not include patients with malaise, fatigue, and nausea in the symptomatic group, while we did. If we had chosen the same group selection criteria as Nyberg & Loof (8), the number of asymptomatic patients would have increased and would probably have changed the survival rate in this group. O n the other hand, like Roll et al. (4), we included in the asympto-
125
matic group two patients with icterus that had not been registered by the patients themselves. Five (38%) of our 13 asymptomatic patients developed symptoms during the time of observation, a percentage similar to that of others (4,8). In a study by James e t al. (10) only 19% of the initially symptom-free patients developed symptoms during the follow-up study. However, the follow-up time was slightly shorter in their study (4.5 years) than in ours (5.7 years). Symptoms developed after a mean of 5 years. O n e may argue that the asymptomatic group of patients probably is part of the same PBC population as the symptomatic patients, and that all of them, if followed long enough, will develop symptoms. This would seem reasonable, considering the poor observed survival rate in the asymptomatic group, although there were few asymptomatic patients in our study, and no absolutc' conclusions with regard to prognostic factors or survival can be drawn. The mean survival rate of 18 years in the symptomatic group is 6 years longer than the best survival rate published earlier, including the recent report from Sweden (8). Using a univariate analysis, we found that there were seven clinical factors that correlated with prognosis: ascites, presence of esophageal varices, jaundice, bleeding from esophageal varices, abdominal pain, hepatomegaly, and age at the time of diagnosis. Roll et al. (4) found icterus, weight loss, hepatomegaly, splenomegaly, and ascites correlating well with shortened survival. Others ( 5 , 8) have also found that hepatomegaly correlated strongly with progression of disease. With regard to the laboratory values in a univariate setup, the logarithms of bilirubin and albumin showed correlation to prognosis, a finding consistent with that of others (4). In accordance with Roll et al. (4), we found that the logarithm of bilirubin correlated with survival and was an independent prognostic factor. Roll (4) also showed that age at the onset of disease and hepatomegaly were independent discriminators of a poor prognosis. Hepatomegaly has also been found by others (8) to be a prognostically ominous sign. In our study gastrointestinal bleeding from esophageal varices correlated to a poor prognosis.
Scand J Gastroenterol Downloaded from informahealthcare.com by University of British Columbia on 10/29/14 For personal use only.
176
A . Ryilning et al.
Other investigators have shown a significant inverse relationship between survival and the extent of fibrosis (4,6,8) and cholestasis (4,6). The histologic data in our study were of no prognostic importance. To our surprise we found a relatively high number of patients with Scheuer grade IIILIV changes among the asymptomatic group. In other studies fibrosis limited to the portal areas indicated a better prognosis (4,s). The differences in prognosis and prognostic factors found in the different studies may partly be due to differences in group criteria and in the nature of the cohorts with regard to the distribution o f risk fiIctors. Another possibility is that there are ethnic differences. Thirdly, there are always weaknesses in retrospective studies, since the amount of accurate information present in the individual case histories and examinations may to some extent vary. However, most of the studies with which we are comparing our results were carried out retrospectively. In conclusion, the survival o f symptomatic patients was longer than previously reported. With reservations because of the small number of asymptomatic patients, their survival seems no better than the survival of the symptomatic group. The presence of variceal gastrointestinal bleeding and the logarithm of bilirubin were found to be independent predictors of mortality. The proportional hazard model applied to a cohort of 52 symptomatic patients with PBC represents a n advanced model for a more precise identification of prognostic variables and provides us with a prognostic forecast for any individual patient. In our opinion it is an important procedure that helps us to identify patients in whom liver transplantation may be necessary. However, studies prospectively examining the clinical and laboratory data are needed to determine whether the prognostic factors maintain their value unchanged over time.
REFERENCES 1. Pugh RNH, Murray-Lyon IM, Dawson JL, Pietroni MC, Williams R. Transection of the oesophagus for bleeding oesophageal varices. Br J Surg lY73, 60. 646-649 2. Sherlock S. Primary biliary cirrhosis (chronic intrahepatic obstructive jaundice). Gastroenterology 1959, 37. 574-558 3. Shapiro JM, Smith H , Schaffner F. Serum bilirubin: a prognostic factor in primary biliary cirrhosis. Gut 1979, 20, 137-140 4. Roll J, Boyer JL, Barry D, Klatskin G . The prognostic importance of clinical and histological features in asymptomatic and symptomatic primary biliary cirrhosis. N Engl J Med 1983, 308. 1-7 5 . Beswick DR, Boyer JL. Primary biliary cirrhosis Hepatology 1984, 4, 29s-33s 6. Christensen E, Crowe J, Doniach D. et al. Clinical pattern and course of disease in primary biliary cirrhosis based on analysis o f 236 patients. Gastroenterology 1980. 78, 236-246 7. Lee RG, Epstein 0, Jauregui H , Sherlock S, Scheuer PJ. Granulomas in primary biliarq cirrhosis: a prognostic feature. Gastroentcrology 1981, 81, 983-986 8. Nyberg A, Loof L. Primary biliary cirrhosis: clinical features and outcome with special reference to asymptomatic disease. Scand J Gastroenterol 19x9, 24, 57-64 9. Shcrlock S, Scheuer PJ. Current concepts. ‘The presentation and diagnosis of 100 patients with primary biliary cirrhosis. N Engl J Med 1973, 289, 671-678 10. James 0, Macklon AF, Watson AJ. Primary biliary cirrhosis-a revised clinical spectrum. Lancet 1081, 1. 1278-1281 11. Scheuer PJ. Liver biopsy interpretation. 4th 4. Bailliere Tindall, London, 1988 12 Breslow N. A generalized Kruskal-Wallis test for comparing K samples subject to unequal patterns censoring. Biometricu 1970, 57, 579-594 13. Mantel N . Evaluation of survival data and new rank order statistic arising in its consideration. Cancer Chemother Rep 1966, 50, 163-170 14. Cox DR. Regression model and lifc table. J K Statistic Soc [B] 1972. 34. 1x7 1s. Balasubramarriam K, Grambsch PM, Wicsner RH, Lindor KD. Dickson ER. Asymptomatic and primary biliary cirrhosis (PBC). Patients have a diminished survival [Ahstritct]. Hepatology 19x7, 7. 1025
Scand J Gastroenterol Downloaded from informahealthcare.com by University of British Columbia on 10/29/14 For personal use only.
A. RYDNING, E. SCHRUMPF, M. ABDELNOOR, K. ELGJO & E. JENSSEN Medical Dept. A and Dept. of Pathology, Rikshospitalet, The National Hospital, and Institute of Medical Statistics, Ullevll Hospital, Oslo, Norway Rydning A, Schrumpf E, Abdelnoor M, Elgjo K, Jenssen E. Factors of prognostic importance in primary biliary cirrhosis. Scand J Gastroenterol 1990, 25, 119-126 To determine survival and the risk factors of death in primary biliary cirrhosis, data from 52 symptomatic and 13 asymptomaticpatients were analyzed. The mean followup time was 6.3 years (range, 0.4-23 years). The average length of survival was 18 years for the symptomaticand 8.4 years for the asymptomaticpatients. By a univariate analysis, ascites, presence of esophageal varices, gastrointestinal bleeding, jaundice, hepatomegaly and the logarithms of albumin and bilirubin were all associated with a poor prognosis. A multivariate analysis of the clinical features showed that the presence of bleeding from esophageal varices and the logarithm of bilirubin were the only predictors for poor prognosis. The survival of the symptomaticpatients is longer than reported previously, while the life expectancy for the asymptomatic patients seems no better than for the symptomatic group. Key words: Clinical features; histology; primary biliary cirrhosis; prognostic factors;
survival Andreas Rydning, M . D . , Medical Dept. A , The National Hospital, 0027 Oslo I , Norway
Primary biliary cirrhosis (PBC) is a chronic cholestatic liver disease of unknown etiology. Without hepatic transplantation the terminal outcome is death in liver failure or bleeding esophageal varices. The prognosis for liver diseases has until recently been assessed largely on the basis of clinical criteria and the use of Child’s classification, often with Pugh’s modification (1). Although serum bilirubin has been shown to be a good marker of progression of disease ( 2 4 ) , other factors have also been shown to be of importance (4,7-9). Today the diagnosis of PBC is often made in asymptomatic patients, probably because of the more elaborate use of liver function tests on routine examinations (10). Symptoms and progressive disease eventually develop in some of these patients. The aim of the study has been to look at survival and determine factors of prognostic importance in patients with symptomatic and asymptomatic PBC. Since liver transplantation is a new thera-
peutic modality in PBC, is is important to determine the natural course of the disease and determine risk factors of mortality, to optimize the timing of transplantation. MATERIALS A N D M E T H O D S During the years 1970-1986, 65 patients with clinical, biochemical, and histologic evidence of PBC were admitted to Medical Dept. A , The National Hospital, Oslo, Norway. A t the time of admission 52 patients (80%) had symptoms of liver disease, whereas 13 (20%) patients were asymptomatic. Patients were regarded and classified as asymptomatic if no symptoms could be attributed to their liver disease. Signs and symptoms more specifically related to PBC include pruritus, icterus, hepatomegaly and/or splenomegaly, and ascites. The non-specific symptoms are malaise o r fatigue, abdominal pain, hyperpigmentation, weight loss, and xanthomas (2-
Scand J Gastroenterol Downloaded from informahealthcare.com by University of British Columbia on 10/29/14 For personal use only.
120
A . Rydning ct ul.
4.6). Other symptoms not related to the liver disease were not included in estimating the start point of the disease. Most of the asymptomatic patients were identified on the basis of a raised serum level of alkaline phosphatase and/or alanine aminotransferase detected during a routine physical examination. Coexisting autoimmune diseases were recorded. In the same period 21 patients with cholestatic !iver disease of uncertain origin were admitted. 'These patients were classified as follows: secondary biliary cirrhosis, cholestatic liver disease, or chronic active hepatitis/PBC?
Nistology Liver biops,y sections were examined by one of [he investigators (K. Elgjo) at the time of initial cv;iluation and re-examined again now. The diagnosis was established and classified in accordance with Scheuer (1 1).
('linical and laboratory data Data on symptoms, physical findings, and laboratory data were obtained from hospital records at the time of diagnosis. The laboratory data analyzed were alkaline phosphatase, alanine aminotransferasc, bilirubin, albumin, prothromhin time. cholesterol, IgM, antinuclear antibody (ANA), smooth-muscle antibody (SMA), and aritimitochondri;rI antibody (AMA). All patients had patent biliary ducts demonstrated by intravenous cholangiography, endoscopic retrograde chol;rngiography, or operative cholangiography. Survioml Information on survival o r death of any patient was obtained from The National Center for Health Statistics. The termination of observation was 1 October 1987. The mean follow-up time was 7 years (range, (1.7-23 years) for the group o f symptomatic patients and 5.7 years (range, 0.4-1 1 . 1 years) for the asymptomatic patients. ,S(afi.sticulanaJysis A preliminary analysis with a univariate method was performed for each postulated progiiostic factor in the symptomatic group of patients, using survival curves. Continuous variables were
tested with the Mann-Whitney or Students' t test. Discrete variables were tested with the contingency table. Comparison of univariate survival curves was made with Breslow (12) and Mantel-Cox test statistics (13). To estimate risk factors for total mortality, the data were analyzed with the Cox regression model (14). The advantages of this model are linkage of survival function to prognostic variables and the possibility of quantifying the effect of prognostic covariates on survival. The essential check of the proportional effect of significant covariate was met by plotting the logarithm of the integrated hazard. The variables estimated by single-factor analysis were subjected to multifactorial analysis by manual backward elimination procedures. Multifactorial procedures were preceded by estimation of correlations between variables. If some variables were significantly correlated, their predictive value was assumed to be similar. The prognostic factors identified as independent were tested for interactions. A BMDP statistical program and an IBM 4341 LO1 computer were used. RESULTS The ratio between male and female patients was similar for both the symptomatic (21 % male, 7Y% female) and the asymptomatic group (15% male, 85% female), and there were no significant differences in age between the two groups at the time of diagnosis (53.1 and 51.3 years, respectively). Our patients were followed up for a mean of 6.3 years, a total of 6.3 X 65 = 409 patient-years. The observed frequency of signs and symptoms is listed in Table I . Symptoms Pruritus was the most frequent initial symptom among the symptomatic patients, experienced in nearly SO% of the cases (Table I). Malaise and abdominal pain were the second and third most frequently registered presenting complaints. Several patients had more than one initial symptom. Five (38%) of the asymptomatic patients developed symptoms during the time of observation (mean, 5 years; range, 2-8 years). Two patients experienced pruritus; one, abdominal pain plus
Prognostic Factors in PBC
121
Scand J Gastroenterol Downloaded from informahealthcare.com by University of British Columbia on 10/29/14 For personal use only.
Table I. Frequency of signs and symptoms in 65 patients with PBC
Symptoms/signs
Symptomatic, % ( n = 52)
Asymptomatic, % ( n = 13)
P
Pruritus Malaise/fatigue Abdominal pain Weight loss Anorexia Nausea Hepatomegaly GI hemorrhage Jaundice Splenomegaly Hyperpigmentation Ascites Xanthomas
49 47 45 37 21 19 53 25 25 17 9 8 4
31
NS
15
NS
8
NS
Table 11. Associated autoimmune disorder in 52 symptomatic and 13 asymptomatic patients with PBC n Thyroid disease Keratoconjunctivitis sicca Scleroderma Disseminated lupus erythematosus Raynaud’s phenomenon Antibody towards salivary gland Congestive cardiomyopathy Rheumatoid arthritis
96
14 5 3 1.5 1.5 1.5 1.5 1.5
malaise; one, ascites, icterus, and dyspnea; and the last patient developed variceal bleeding and coma.
Signs A m o n g the physical signs registered at the time of diagnosis, hepatomegaly was by far the most
frequent sign seen both among the symptomatic (53%) and the asymptomatic (31%) patients (Table I). A t diagnosis of PBC gastrointestinal hemorrhage was seen in 25% of symptomatic and in none of the asymptomatic patients, while jaundice was equally present in the two groups.
Autoimmune phenomena Autoimmune phenomena were present in 17 patients (12 symptomatic and 5 asymptomatic) (Table 11). Three patients had more than one autoimmune phenomenon. Laboratory values (Table I l l ) Mean level of alkaline phosphatase was 675.1 U/1 in the symptomatic and 830.4 U/1 in the asymptomatic group. T h e same tendency with higher levels in asymptomatic patients was seen with regard to alanine aminotransferase ( A L A T ) . Mean serum bilirubin was 29.9 pmol/l in the
Table 111. Laboratory data in 65 patients with PBC Symptomatic
Alkaline phosphatase (55-195 U/1) Alanine aminotransferase (9-40 U/I) Bilirubin (3-26 pmol/l) Albumin (35-50 g/l) Prothrombin time (65-150%) Cholesterol (2.S6.5 mmol/l) IgM (0.5-2.8 g/l)
Asymptomatic
n
Mean
SD
n
Mean
SD
52 52 52 51 51 39 49
676 75 30 34 101 7.2 6.4
539 58 34 8.7 33 3.2 5.1
13 13 13 13 10 11 12
830 96 17 30 118 7.5 7.0
455 80 12 17 34 3.7 3.8
122
A. Kydning et ul.
Tahlc IV. Percentage of positive autoantibodies in 65 patients with PBC I _
Autoantibody
Symptomatic ( n = 52)
Asymptomatic (n = 13)
96 42 2s 9
92 50 46 15
Scand J Gastroenterol Downloaded from informahealthcare.com by University of British Columbia on 10/29/14 For personal use only.
Mitochondria1 antibody l i t e r 2 512 Smooth-muscle antibody Antinuclear antibody
1';rhlc V. Histologic grading o f the first representative liver biopsy in symptomatic and asymptomatic PBC Symptomatic ( n = 52)
Iflstoioglc stage
Asymptomatic ( n = 13)
P NS NS NS
NS
A M A were positive in nearly all o f the patients regardless of symptoms (Table IV). The numbers of patients with SMA and A N A were similar in the two groups of patients.
I I1
4 35 1
symptomatic and 16.6 pinol/l in the asymptomatic group. None of the differences above reached statistical significance. Nor were there any statistical differences between the two groups in levels of albumin, prothrombin time, total cholesterol, and total IgM. Equal numbers of patients had a raised IgM level (IgM > 2.8 g/l) in the symptomatic and the asymptomatic group.
Operations Twelve symptomatic patients were operated on with portosystemic shunts during the follow-up period. Ten patients received a portacaval shunt and two patients received a splenorenal shunt. Four symptomatic patients received a liver trnnsplant. They were followed up until the time when they received a new liver. Histology (Table V ) Among the 52 symptomatic patients there was 1 patient from whom we did not obtain a biopsy, 1patient with normal histology, and 35 with histo-
% Survival
30 -
-
20 10
m.
-
I
7-m
I
2
i
4
I
6
I
8
1
I
i
0
15
m m
Nomslpapul Symptomatic A$ymplomatii
I
I
2o
Years
25
Fig. I . Survival curves for the 52 symptomatic and the 13 asymptomatic patients with PBC.
Prognostic Factors in PBC
123
factors. The presence of ascites resulted in a lower 5-year survival (25%) than in patients without this sign (85%) (Table VI). Patients with varices had a pool survival of 44% versus 88% for patients without. Jaundice gave a pool survival of 75% versus 88%. The presence of gastrointestinal bleeding caused a lower survival rate, 54%, versus 90% for those without this sign. Patients without abdominal pain seemed to have a lower survival Survival Survival curves were drawn for the symp- rate than patients who did experience pain (72% tomatic and the asymptomatic group as shown in versus 92%). Hepatomegaly resulted in poorer Fig. 1. There were 17 (33%) out of 52 PBC- survival, 74% versus 88% in those with a normalrelated deaths in the symptomatic group. Survival sized liver. The other dichotomized variablestimes were calculated from the time of onset splenomegaly , presence of xanthomas, presence of symptoms until death. The observed mean of anorexia, nausea, hyperpigmentation, pressurvival time was 18 years for the symptomatic ence of itching, and loss of weight-did not influgroup. There were five deaths not related to liver ence survival prognosis. The liver histologic data did not influence survival either. disease. A two-sample t test o r the Mann-Whitney test In the asymptomatic group four (31%) deaths were related to the liver disease. O n e additional was carried out, comparing the continuous varideath was not related to liver disease. Survival ables, depending on whether the distribution of times for this group of patients were calculated the variables was normal o r skewed. Age at diagfrom the time of diagnosis, which usually took nosis came out as a significant factor (Table VII). place at their first hospital admission. The After studying the distribution of these factors, observed mean survival for this group of patients we performed a logarithmic transformation, since most of the values had a high degree of skewness. was 8.4 years (Fig. 1). The logarithms of albumin and bilirubin were both significantly different among patients who Univariate analysis Patients who survived were compared with died and patients who are still alive (Table VII), patients who died, with regard to the following whereas the logarithms of alkaline phospatase,
Scand J Gastroenterol Downloaded from informahealthcare.com by University of British Columbia on 10/29/14 For personal use only.
logic changes in accordance with Scheuer grades 111 and IV. Among the 13 asymptomatic patients there were 9 with Scheuer grade I11 and IV changes and only 3 with grade I and I1 changes. In one patient the histologic material was insufficient for evaluation.
Table VI. Clinical factors of prognostic importance for survival (univariate analysis) in 52 symptomatic PBC patients Sign/symptom (variable) Ascites Varices
GI bleeding Jaundice Abdominal pain Hepatomegaly
Cumulative 5-year survival -+ SEM
Chi-square Breslow (PI
Chi-square Mantel-Cox (P)
15.4 (0.0001)
17.6 (0.0001)
11.6 (0.0001)
11.8 (0.0001)
43
25 2 0.22 85 t 0.05 44 5 0.17 88 2 0.05
39
54 2 0.14 90 -t 0.05
9.7 (0.002)
9.0 (0.003)
25
75 2 0.15 88 2 0.07
7.4 (0.007)
11.4 (0.0007)
92 2 0.06 71 2 0.09 74 -t 0.09 88 2 0.07
5.8 (0.015)
4.7 (0.03)
4.1 (0.04)
4.4 (0.03)
Present, yes/no
4 48 9 13 8 24 28 27 25
I'ithle VII. (.'linical and laboratory values of prognostic importance (continuous variables) in 52 symptomatic puticnts
Alive
v.(11."i'ible (log)
n
Mean
SD
n
Mean
SD
I'
Agc at diagnosis
3.5 34 35
1.72 1 .so 1.20
0.0')
17 17 17
1.77 I .4X 1 .55
0.09
0.009 0.002 0.005
Albumin l3ilirul)in Scand J Gastroenterol Downloaded from informahealthcare.com by University of British Columbia on 10/29/14 For personal use only.
Dead
0.26
0.34
0.08 0.29
i';ihle VI11. Results of thc inultivariatc Cox modcl analysis: risk factors of mortality ~~
~
Variables Presence of bleeding 1-og hilirubin
Level
73
SE(t%
Y es/no
1.6812 2.0274
0.5577 0.7 101
~
.41..4'1~.prothrombin time, and cholesterol were not. Another univariate analysis was performed rising t h c C o x inodel and the logarithms of the continuous variables. Among the continuous variables. the logiarithm of bilirubin was highly \igi.rificant (p = O.Ol), whereas the logarithm of AIAT did n o t reach statistical significance ( p = 0.09).
Mitltiouriute irnidy.si.s After ;I tinivariate analysis of the different postulated prognostic factors had been completed, these I ? covariates were pooled together in a multivariate an;ilysis by the proportional hazard iiiodel. A multivariate backward elimination proivtitire was u:;ecl. After different steps estimating thc Iikclihood of the model, seven steps were run until the most stable model was established. The
&S E (
p)
3.0 145
2.8551
presence of gastrointestinal bleeding and the logarithm of bilirubin proved to be independcrit predictors of mortality. The model for the relative risk o f mortality (RR) (Table VIII) in patients with PBC was derived as follows:
RR
= (,I.h8I?(hkcdlng
(I.:?)
+ 2,(i771~,1,ihlllrllhln
DISCUSSION The patients included in this retrospective study had been admitted from all of Norway. although most of them were from the southeast part of the country.
Prognostic I'rcsencc of G I bleeding
Case A
Case B
Case C
Case D
NO
YES
NO
YES
31.3
31.3
0.65
3.52
('oncentration of bilirubin (pmol/l)
100
100
liclativc risk 01 mortality of I'BC
31) i l
By means of this model, the relative risk5 for four hypothetical patients have heen forecast ('Table IX).
I ;ibk IM. Relative risk of mortality Eor lour patients with different subsets of prognostic variables
variable
~
27.5
147
Scand J Gastroenterol Downloaded from informahealthcare.com by University of British Columbia on 10/29/14 For personal use only.
Prognostic Factors in PBC
Nyberg & Loof (8) found that 70% of their 80 patients were asymptomatic at initial presentation, in contrast to the report from Sherlock & Scheuer (9) in 1973, which described the initial presentation of PBC as asymptomatic in 11% of the cases. In our study 20% of patients were initially asymptomatic. The difference in ratio between asymptomatic and symptomatic patients in these studies is most likely related to the period of inclusion. The Swedish study is a follow-up from the years 1979-1986, when more elaborate laboratory data and health investigations were performed, while our first inclusion was in 1970. Others ( 4 , 5 , 9 ) have confirmed that there has been a shift from symptomatic to asymptomatic PBC during the past decade, most likely due to the increased use of laboratory profiles. Several authors have suggested a better prognosis for those patients who are asymptomatic at the time of diagnosis. Roll et al. (4) and Beswick Kr Boyer (5) found a survival rate for asymptomatic patients equal to that of an age- and sexmatched cohort of the general population. Nyberg & Loof (8) report a 50% survival of 16 years in asymptomatic PBC, the reduced survival rate appearing more marked after 12 years of observation. Balasubramarriam et al. (15) published similar poor survival figures for asymptomatic patients, with a drastic increase in mortality appearing after 5 years. In our limited number of asymptomatic patients the survival rate was no better than for the symptomatic patients. Survival may be influenced by the difference in the starting point for registration in the two groups, since the asymptomatic group was registered from the time of diagnosis, and the symptomatic group was registered from the time the symptoms started. Differences in survival times reported between different authors may be due to different criteria chosen for symptomatic/asymptomatic PBC. Nyberg & Loof (8) d o not include patients with malaise, fatigue, and nausea in the symptomatic group, while we did. If we had chosen the same group selection criteria as Nyberg & Loof (8), the number of asymptomatic patients would have increased and would probably have changed the survival rate in this group. O n the other hand, like Roll et al. (4), we included in the asympto-
125
matic group two patients with icterus that had not been registered by the patients themselves. Five (38%) of our 13 asymptomatic patients developed symptoms during the time of observation, a percentage similar to that of others (4,8). In a study by James e t al. (10) only 19% of the initially symptom-free patients developed symptoms during the follow-up study. However, the follow-up time was slightly shorter in their study (4.5 years) than in ours (5.7 years). Symptoms developed after a mean of 5 years. O n e may argue that the asymptomatic group of patients probably is part of the same PBC population as the symptomatic patients, and that all of them, if followed long enough, will develop symptoms. This would seem reasonable, considering the poor observed survival rate in the asymptomatic group, although there were few asymptomatic patients in our study, and no absolutc' conclusions with regard to prognostic factors or survival can be drawn. The mean survival rate of 18 years in the symptomatic group is 6 years longer than the best survival rate published earlier, including the recent report from Sweden (8). Using a univariate analysis, we found that there were seven clinical factors that correlated with prognosis: ascites, presence of esophageal varices, jaundice, bleeding from esophageal varices, abdominal pain, hepatomegaly, and age at the time of diagnosis. Roll et al. (4) found icterus, weight loss, hepatomegaly, splenomegaly, and ascites correlating well with shortened survival. Others ( 5 , 8) have also found that hepatomegaly correlated strongly with progression of disease. With regard to the laboratory values in a univariate setup, the logarithms of bilirubin and albumin showed correlation to prognosis, a finding consistent with that of others (4). In accordance with Roll et al. (4), we found that the logarithm of bilirubin correlated with survival and was an independent prognostic factor. Roll (4) also showed that age at the onset of disease and hepatomegaly were independent discriminators of a poor prognosis. Hepatomegaly has also been found by others (8) to be a prognostically ominous sign. In our study gastrointestinal bleeding from esophageal varices correlated to a poor prognosis.
Scand J Gastroenterol Downloaded from informahealthcare.com by University of British Columbia on 10/29/14 For personal use only.
176
A . Ryilning et al.
Other investigators have shown a significant inverse relationship between survival and the extent of fibrosis (4,6,8) and cholestasis (4,6). The histologic data in our study were of no prognostic importance. To our surprise we found a relatively high number of patients with Scheuer grade IIILIV changes among the asymptomatic group. In other studies fibrosis limited to the portal areas indicated a better prognosis (4,s). The differences in prognosis and prognostic factors found in the different studies may partly be due to differences in group criteria and in the nature of the cohorts with regard to the distribution o f risk fiIctors. Another possibility is that there are ethnic differences. Thirdly, there are always weaknesses in retrospective studies, since the amount of accurate information present in the individual case histories and examinations may to some extent vary. However, most of the studies with which we are comparing our results were carried out retrospectively. In conclusion, the survival o f symptomatic patients was longer than previously reported. With reservations because of the small number of asymptomatic patients, their survival seems no better than the survival of the symptomatic group. The presence of variceal gastrointestinal bleeding and the logarithm of bilirubin were found to be independent predictors of mortality. The proportional hazard model applied to a cohort of 52 symptomatic patients with PBC represents a n advanced model for a more precise identification of prognostic variables and provides us with a prognostic forecast for any individual patient. In our opinion it is an important procedure that helps us to identify patients in whom liver transplantation may be necessary. However, studies prospectively examining the clinical and laboratory data are needed to determine whether the prognostic factors maintain their value unchanged over time.
REFERENCES 1. Pugh RNH, Murray-Lyon IM, Dawson JL, Pietroni MC, Williams R. Transection of the oesophagus for bleeding oesophageal varices. Br J Surg lY73, 60. 646-649 2. Sherlock S. Primary biliary cirrhosis (chronic intrahepatic obstructive jaundice). Gastroenterology 1959, 37. 574-558 3. Shapiro JM, Smith H , Schaffner F. Serum bilirubin: a prognostic factor in primary biliary cirrhosis. Gut 1979, 20, 137-140 4. Roll J, Boyer JL, Barry D, Klatskin G . The prognostic importance of clinical and histological features in asymptomatic and symptomatic primary biliary cirrhosis. N Engl J Med 1983, 308. 1-7 5 . Beswick DR, Boyer JL. Primary biliary cirrhosis Hepatology 1984, 4, 29s-33s 6. Christensen E, Crowe J, Doniach D. et al. Clinical pattern and course of disease in primary biliary cirrhosis based on analysis o f 236 patients. Gastroenterology 1980. 78, 236-246 7. Lee RG, Epstein 0, Jauregui H , Sherlock S, Scheuer PJ. Granulomas in primary biliarq cirrhosis: a prognostic feature. Gastroentcrology 1981, 81, 983-986 8. Nyberg A, Loof L. Primary biliary cirrhosis: clinical features and outcome with special reference to asymptomatic disease. Scand J Gastroenterol 19x9, 24, 57-64 9. Shcrlock S, Scheuer PJ. Current concepts. ‘The presentation and diagnosis of 100 patients with primary biliary cirrhosis. N Engl J Med 1973, 289, 671-678 10. James 0, Macklon AF, Watson AJ. Primary biliary cirrhosis-a revised clinical spectrum. Lancet 1081, 1. 1278-1281 11. Scheuer PJ. Liver biopsy interpretation. 4th 4. Bailliere Tindall, London, 1988 12 Breslow N. A generalized Kruskal-Wallis test for comparing K samples subject to unequal patterns censoring. Biometricu 1970, 57, 579-594 13. Mantel N . Evaluation of survival data and new rank order statistic arising in its consideration. Cancer Chemother Rep 1966, 50, 163-170 14. Cox DR. Regression model and lifc table. J K Statistic Soc [B] 1972. 34. 1x7 1s. Balasubramarriam K, Grambsch PM, Wicsner RH, Lindor KD. Dickson ER. Asymptomatic and primary biliary cirrhosis (PBC). Patients have a diminished survival [Ahstritct]. Hepatology 19x7, 7. 1025
