Thrombocytopenia in liver disease

EPIDEMIOLOGY OF THROMBOCYTOPENIA IN LIVER DISEASE Thrombocytopenia in cirrhosis of the liver is a fairly frequently observed phenomenon, occurring in 15% to 70% of patients with cirrhosis, depending on the stage of disease and the definition of thrombocytopenia (1). It is usually mild to moderate in severity, and its degree can be viewed as a prognostic marker (2). However, there is little evidence to date suggesting an association with the increased risk of bleeding from esophageal varices or other causes in patients with stable liver disease (3-5), unless undergoing major surgery (6). Even liver biopsy appears to be quite safe unless severe

Thrombocytopenia in liver disease can have various etiologies.Mostly, thrombocytopenia is associated with liver cirrhosis but can also be due to autoantibodies against platelets (9,10), the myelosuppressive action of viral agents such as hepatitis C virus (HCV) (11,12), the toxic effects of excessive alcohol ingestion and therapy of chronic liver disease with interferon (13,14).This review will focus on the pathophysiology of thrombocytopenia associated with cirrhosis of the liver; platelet autoantibodies will not be discussed in detail.

THROMBOCYTOPENIA IS LINKED TO CIRRHOSIS
Thrombocytopenia in liver disease is always linked to cirrhosis, except in cases of HCV infection, where either autoantibodies against platelet antigens, or the HCV itself, can cause thrombocytopenia in noncirrhotic patients through bone marrow suppression (12).
The mechanisms leading to thrombocytopenia in cirrhosis of the liver, aside from HCV infection, are not so clear.Most commonly, thrombocytopenia has been attributed to portal hypertension and splenic pooling of platelets (15).Low-level consumption coagulopathy has been discussed in relation to cirrhosis and has been suggested as being a possible cause of thrombocytopenia, but careful coagulation studies in patients with cirrhosis do not suggest consumption coagulopathy in most patients with cirrhosis (16).Furthermore, data from elegant studies, such as those by Violi et al (8,17), do not show lower platelet counts in patients with low-level disseminated intravascular coagulation compared with patients without the condition.
Recently, decreased thrombopoeitin (TPO) production in the cirrhotic liver has been show to play an important role and may be the most important factor for the majority of patients with cirrhosis and thrombocytopenia (18)(19)(20).

PORTAL HYPERTENSION -
THE EVIDENCE For many years, portal hypertension was thought to induce splenomegaly, and splenomegaly was presumed to cause thrombocytopenia through an increased pooling or destruction of platelets (15).Splenic pooling of platelets has been demonstrated in experimental studies in normal, splenectomized and hypersplenic animals without portal hypertension, where platelet counts correlated with spleen weight and total body platelet mass was similar in the three groups (21).Surprisingly, platelet survival time was also the same in all three groups, because platelet removal in splenectomized animals was largely accomplished by the liver.
Human cases were reported by Aster (15) and Penny et al (22), who recovered 62% to 89% of total body platelet mass after ex vivo perfusion of massively enlarged spleens in one patient with cirrhosis and one with extraportal hypertension, respectively.
The clinical data on relief of portal hypertension do not show such unequivocal results.Several studies report on the effect of portocaval shunts on hypersplenism in cirrhosis of the liver (for an incomplete list see Table 1).
After portocaval shunt operations for variceal bleeding, Rousselot et al (23) could not demonstrate a significant improvement in platelet counts, while Hutson et al (24) claimed that his group did.In fact, there was improvement in peripheral platelet count after shunt surgery in 11 of 13 thrombocytopenic patients, but many of them still remained thrombocytopenic.In addition, the analysis did not take into account that six patients with a normal platelet count before surgery became thrombocytopenic after shunt surgery.Overall, there was no significant difference between platelet counts before and after shunt surgery.
Almost all of these studies suffer from severe methodological problems.They are retrospective analyses of patients who underwent shunt surgery for bleeding varices at various time points after the bleeding episode.It is impossible to distinguish between spontaneous recovery of platelet count after a bleeding episode, which often times is quite dramatic, and the true effects of portal decompression on the incidence of 'hypersplenism'.None of these studies contain a control group of patients undergoing conservative or endoscopic treatment, and most of them only report on a very select segment of their patient population undergoing shunt surgerynamely those with pronounced thrombocytopenia but relatively good liver function (allowing them to undergo shunt surgery).In essence, no conclusion can be drawn from these studies regarding the effect of portocaval anastomosis on peripheral platelet counts in patients with cirrhosis of the liver.
The exception is a study by Mutchnick et al (25), who reported on patients with portal hypertension and esophageal varices but without a prior bleeding episode.The patients were placed randomly in controlled trials of prophylactic shunt surgery.Portal pressure gradient, and incidence of splenomegaly and thrombocytopenia were similar in both groups.While the reversal of splenomegaly in the shunt group did not quite reach statistical significance (after an average follow-up of 5.5 years), the effects on thrombocytopenia were the same in both groups.No improvement of thrombocytopenia could be found in either group.If anything, the incidence of thrombocytopenia increased in both the control group and surgical shunt group during the five years of follow-up.
Several studies of thrombocytopenia post-transjugular intrahepatic portosystemic shunt (TIPS) have been published, but most of them report on use only after an acute bleeding episode or on selected patients with more severe thrombocytopenia, show only limited data, and suffer from the same shortcomings as the surgical shunt studies (26)(27)(28)(29).A study of 44 unselected patients by the author showed no increase in platelet count, neither at 14 days nor at one year post-TIPS implantation (30), a finding also observed by others (31).
The correlation between spleen size and platelet count in patients with cirrhosis is also not firmly established.Jalan et al (28) showed that, in patients who underwent TIPS placement, there was an improvement in platelet count, but failed to show that there was a correlation between either portal pressure or spleen size and the recovery of platelet count.Data from the author (unpublished observations) on patients undergoing liver transplantation also do not indicate any correlation between spleen size and platelet count, while Mutchnick et al (25) in their study on portocaval anastomosis patients at least noted a trend.
Several groups have demonstrated the recovery of platelet counts in thrombocytopenic patients with cirrhosis of the liver after embolization of large parts of the spleen (32)(33)(34).In principle, this is another form of splenectomy and increases peripheral platelet count, most likely by prolonging platelet survival time.In animal studies of experimental splenectomy, platelet survival was unchanged in both splenectomized and control rats because the liver took over platelet removal from the spleen.This might not be the case in humans, however, especially when the liver is cirrhotic.
There are very few studies on peripheral platelet counts in noncirrhotic patients with extrahepatic or presinusoidal portal hypertension.The effects of extrahepatic portal hypertension on peripheral platelet counts have been studied in patients with portal venous thrombosis (35), while presinusoidal hypertension without severe hepatocellular damage can be found in patients with schistosomiasis (36).Platelet counts in patients with extrahepatic portal hypertension were between normal controls and patients with portal hypertension and liver disease (35), suggesting dual mechanisms for thrombocytopenia in liver disease: splenic sequestration of platelets due to portal hypertension and a further decrease in peripheral platelet count due to reduced hepatocellular function.This is also demonstrated by heterotopic liver transplantation (HLT), which relieves portal hypertension by the creation of a portocaval shunt and restores liver synthetic function by transplantation of a functioning liver graft.HLT resolves thrombocytopenia very effectively (37), contrary to the simple relief of portal hypertension without restoration of liver synthetic function.Therefore, portal hypertension seems to play some role in thrombocytopenia in liver disease, but adequate liver synthetic function is essential for the maintenance or restoration of a normal peripheral platelet count.

BONE MARROW PRODUCTION OF PLATELETS IN CIRRHOSIS
Indirect information about bone marrow production of thrombocytes is available from various turnover studies, but the methodology used in the studies was quite heterogeneous as are the results of these studies.Different radioligands ( 51 Cr, 111 In) used for labelling of platelets and different methodologies for calculating platelet turnover (38) make it difficult to compare these studies.Very little direct information is available about the bone marrow of patients with cirrhosis of the liver.After using commonly available search machines and looking through citations in older literature extending beyond the 1960s, no systematic study on megakaryopoeisis and thrombocytopoiesis in patients with cirrhosis of the liver could be found, probably because quantification of megakaryopoiesis is methodologically cumbersome and biopsy of bone marrow is rarely justified by the moderate degree of cytopenia.One study by Harker and Finch (39) assessed thrombokinetics in 83 patients with thrombocytopenia, including one patient with cirrhosis of the liver, and found that thrombocyte production decreased to 70% of normal.Another study by the same author (40) showed a wide range of platelet production rates, platelet survival and peripheral platelet counts in a very heterogeneous population of patients with cirrhosis of the liver.In these patients, the peripheral platelet count was a direct manifestation of platelet production in the bone marrow.In a recent study using reticulated platelets in the peripheral blood as markers of thrombopoiesis, it could be demonstrated that thrombocytopenia in liver disease is caused by a low platelet production rate, just as in aplastic anemia and unlike thrombocytopenia with high platelet turnover in conditions such as idiopathic thrombocytopenic purpura (ITP) (41).

THROMBOPOIETIN
In 1994, TPO was discovered.TPO is almost exclusively expressed in the liver, with only minor amounts originating from the kidneys and other sources (42).What makes TPO so ideal for playing a key role in the etiology of thrombocytopenia in liver disease is the unique mechanism of regulating TPO serum levels.TPO serum levels depend on two major factors: TPO production in the liver; and the binding of TPO to its receptors, which are located on thrombocytes and megakaryocytes (43).TPO clearance from the circulation depends on peripheral platelet count.Even though megakaryocytes contain up to 10 times more receptors per cell than circulating platelets, the TPO receptors found on platelets still make up more than 95% of the TPO-removing receptors.TPO production by the liver is constant and is not regulated at the transcriptional level (44).One hepatocyte seems to produce the same amount of TPO at all times, and upregulation of TPO production does not occur.Therefore, the amount of TPO produced in the liver depends on the functional liver cell mass, providing the rationale to assume that TPO plays a role in thrombocytopenia in liver disease.

EXPERIMENTAL EVIDENCE
In 1975, Siemensma and colleagues (45) showed in an animal model that peripheral platelet count is somewhat related to functional liver mass.After a four-fifths hepatectomy, animals became thrombocytopenic within two days and their peripheral platelet counts recovered only by day 8, when the liver had already regained 80% of its previous weight.In the same study, they showed that recovery from acute thrombocytopenia induced by treatment with antiplatelet serum took much longer in rats after a four-fifths he-

62D
Can J Gastroenterol Vol 14 Suppl D November 2000

Peck-Radosavljevic
patectomy than in control animals, again linking peripheral platelet count to liver cell mass.
After the cloning of TPO and its receptor (c-mpl), it was demonstrated in knock-out animals that there was a clear dose-response relationship between either TPO or its receptor c-mpl and peripheral platelet counts.Homozygous TPO knock-out mice showed a greater than 80% reduction in peripheral platelet mass, while heterozygous animals had 60% of the peripheral platelet count of control animals (46).Likewise, c-mpl knock-out mice had a greater than 85% reduction of peripheral platelet count (47).In very elegant experiments, Qian et al (48) tried to determine the contribution of liver-derived TPO to megakaryopoiesis and thrombocytopoiesis in mice by transplanting livers from TPO knock-out mice into syngeneic wild-type mice.They showed that the liver contributes 60% of the TPO required for maintenance of normal platelet count in mice.In humans, this fraction will likely be higher because, as judged from mRNA expression in Northern blots, TPO in humans is mostly produced in the liver, and TPO production in the kidneys of mice exceeds that in humans markedly (42,49).

PATIENTS WITH CIRRHOSIS OF THE LIVER
Several groups have published data on TPO serum or plasma levels in patients with cirrhosis of the liver and have come to opposing conclusions in the interpretation of their data.Overall, most groups have found unchanged, slightly elevated or reduced TPO levels in patients with various stages of liver cirrhosis compared with control patients (18)(19)(20)(50)(51)(52)(53)(54)(55)(56)(57).All of the reported TPO levels in patients with cirrhosis of the liver were in the range of normal.Even though the authors have tried to come to different from their data, little insight can be gained without additional information about platelet production and turnover in these patients.The only mechanism for thrombocytopenia in liver disease that can be excluded from these data is predominant bone marrow failure due to liver disease-associated toxins with normal TPO production by the liver.This would require high TPO serum levels in thrombocytopenic patients, similar to patients undergoing chemotherapy for malignant disease, but high TPO levels are not observed in patients with liver disease.
The author and others have demonstrated that TPO serum levels are in the normal range in patients with advanced-stage liver disease and thrombocytopenia.TPO levels in the normal range with thrombocytopenia cannot be regarded as 'normal', because a compensatory increase in TPO serum levels would be expected (41,50,57,58).Normal TPO levels with low peripheral platelet count can theoretically be observed in two clinical conditions: in high turnover states such as ITP, where bone marrow production of platelets is increased but newly formed platelets are rapidly destroyed, or through reduced TPO production by the liver, which usually does not occur.Controversy about the pathophysiology of thrombocytopenia in liver disease exists because in the past, a state of consumption coagulopathy has been postulated for patients with advanced-stage liver disease.Overt consumption coagulopathy can only be observed during episodes of active bleeding or in acute liver failure (59).Low-level consumption coagulopathy can sometimes also be observed in patients with stable liver disease.It seems to occur mainly during periods of endotoxemia and affects the plasmatic coagulation system but has not been shown to have any influence on peripheral platelet count (17).Overall, consumption coagulopathy, even at low levels, seems to be more the exception than the rule, even in advanced-stage liver disease (16).
Recently, a study published by Koike et al (41) was able to shed some light onto this issue.By comparing platelet production (measured as reticulated platelets) and TPO serum levels in thrombocytopenic patients with aplastic anemia, ITP and cirrhosis of the liver, they were able to show that TPO levels are increased and platelet production is decreased in patients with aplastic anemia.Patients with cirrhosis show normal TPO levels, which are too low considering the peripheral platelet count in these patients, while platelet production by the bone marrow is reduced.This is the first study showing that the gradual decline in liver function in cirrhosis, with a gradual decline in TPO production by the liver, results in a low platelet production Can J Gastroenterol Vol 14 Suppl D November 2000 63D Thrombocytopenia in liver disease

PATIENTS UNDERGOING LIVER TRANSPLANTATION
Because the serum levels of TPO in patients with cirrhosis do not explain the mechanism leading to thrombocytopenia in cirrhosis of the liver, the next area of study was liver transplantation.It had been shown that liver transplantation, either orthotopic liver transplantation (OLT) (1) or HLT (37), is associated with a rapid correction of thrombocytopenia if present before liver transplantation.It was hypothesized that TPO serum levels and a declining peripheral platelet count during the progression of chronic liver disease always remained in an equilibrium, characterized by reduced platelet production and consequently less removal of TPO from the circulation by binding.Liver transplantation in patients with advanced-stage liver disease and thrombocytopenia would restore normal TPO production immediately with a persistently low platelet count.This would lead to a sharp rise in TPO serum levels within the first two days of liver transplantation.Increased TPO serum levels post-OLT would only decline together with a marked increase in peripheral platelet count.
Indeed, it could be shown that TPO levels start to increase at day 1 post-transplantation and that the increase in TPO levels paralleled the course of recovery of liver synthetic function as shown by the increase in clotting factors (18,58).The increase in TPO serum levels is dependent on the pre-OLT platelet count, because only patients with thrombocytopenia before OLT show an increase in TPO serum levels, while patients with normal platelet count before OLT do not (50).The gradual increase in TPO serum levels and clotting factors within the initial two to three days after OLT did not come as a surprise, because immediately after reperfusion of the graft, coagulation and hyperfibrinolysis activation are well documented (60,61).This hypercoagulative state causes consumption of clotting factors and thrombocytes, but subsides after the first day after surgery.Together with thrombocyte consumption on day 1, the clearance of TPO from the circulation is accelerated.Nevertheless, TPO serum levels are significantly elevated already on day 1 post-OLT and peak between days 2 to 5 post-OLT when platelet counts are still low without any ongoing coagulation activation or platelet consumption (58).Around day 6 post-OLT, the peripheral platelet count starts to increase, which is in good accordance with the time lag observed between the administration of recombinant human TPO and the increase in peripheral platelet count in humans (62).By day 14, most patients have peripheral platelet counts in the normal range and all patients with an uncomplicated postoperative course have significantly higher platelet counts than before OLT (18).This increase in peripheral platelet count does not simply reflect relief of portal hypertension, which is a consequence of OLT, but is related to liver graft function (61,63,64).
This relation to liver graft function is supported by the fact that the increase in platelet count post-OLT is selective for thrombocyte counts, as expected for the effect exerted by a lineage-specific growth factor like TPO.Neither hemoglobin nor white blood cells show any significant change within the first 14 days post-OLT, as one would expect if relief of 'hypersplenism' was the mechanism for the increase in peripheral platelet count post-OLT.

CONCLUSIONS
The etiology of thrombocytopenia in liver disease cannot be attributed to a single cause with certainty.Even after decades of holding 'hypersplenism' as the major theory to explain low peripheral platelet counts in patients with liver cirrhosis, clarifying studies showing the correlation between portal pressure, spleen size, bone marrow production of platelets, platelet survival time and peripheral platelet count are lacking.The evidence to date in the etiology of thrombocytopenia in liver disease favours a combined role of portal hypertension and splenic sequestration on one hand, and of decreased TPO production by the diseased liver on the other hand.This combination is suggested because decreased TPO production on its own cannot explain thrombocytopenia in liver disease in all patients convincingly.
The ultimate proof of the role of portal hypertension versus the lack of TPO production in thrombocytopenia would be the correction of thrombocytopenia by the administration of recombinant human TPO (rH-TPO) (Genentech Inc, USA) or its synthetic C-terminally truncated and pegylated analogue, recombinant human megakaryocyte growth and development factor (rH-MGDF) (Amgen Inc, USA).Unfortunately, two phase 1 studies of patients with cirrhosis and thrombocytopenia were stopped before they were started due to problems with the recombinant TPO analogue rH-MGDF.The analogue caused production of neutralizing antibodies that cross-reacted with the endogenous molecule in healthy individuals after repeat administration in platelet donors (65).Problems with neutralizing antibodies have not been described in several hundred oncological patients receiving rH-MGDF (66,67) or the full-length recombinant molecule rH-TPO (62), maybe because these patients were immunosuppressed to a certain degree.Otherwise, the safety profile of the recombinant TPOs are very favourable, with virtually no side effects except for very high peripheral platelet counts (greater than 1,000,000/µL) in some patients, which have not caused any thrombotic complications so far.In particular, platelet aggregation and activation are not induced through rH-TPO (68).After careful assessment of the problem with neutralizing antibodies, clinical studies involving the various states of thrombocytopenia, outside the oncological setting, should resume.Patients with liver disease would be an ideal target population for these studies, once phase 1 studies in patients with stable cirrhosis and thrombocytopenia have shown safety and efficacy.Possible indications to study the efficacy of rH-TPO would be thrombocytopenic patients undergoing invasive diagnostic procedures or surgery, patients after variceal hemorrhage, patients undergoing OLT to decrease the rate of postoperative bleed-

64D
Can J Gastroenterol Vol 14 Suppl D November 2000

Peck-Radosavljevic
ing complications or patients awaiting OLT to decrease the intraoperative need for blood products.Maybe the most interesting patient group to study would be patients undergoing interferon-alpha treatment for chronic hepatitis C. Thrombocytopenia is a frequent and dose-dependent side effect of interferon treatment for chronic hepatitis C and is already the reason for a dose reduction or termination of treatment in a considerable number of patients.As shown recently, not only bone marrow production of platelets, but also TPO production in the liver, is hampered by interferon therapy (20).There is increasing evidence that higher doses and prolonged treatment intervals might improve virological responses (69), and that the treatment of patients with hepatitis C cirrhosis (previously not considered a group to benefit from interferon therapy) might improve their survival (70,71) and reduce the incidence of hepatocellular carcinoma (71,72).Therefore, a potent yet well-tolerated thrombopoietic agent will be in high demand, an agent such as recombinant human TPOs.It is hoped that it will be available to treat thrombocytopenia in liver disease in the future.