In 1980, we reported that patients receiving high-dose cytoreductive therapy had a higher incidence of VOD than did those receiving low-dose therapy [7]. In recent years, the trend has been to give even greater doses of cytoreductive therapy to patients at increased risk for relapse of malignancy after marrow grafting [5, 8-10]. High-dose regimens reduce the incidence of tumor relapse but are associated with liver, renal, cardiac, and pulmonary complications [8-12]. Our clinical impression is that the current incidence of VOD at our institution is much higher than the 21% rate we reported 9 years ago [2] and that more patients have severe liver disease.

Factors other than high-dose cytoreductive therapy must be involved in the pathogenesis of VOD, because not all patients receiving identical high-dose regimens develop this complication. In three studies, patients with hepatitis at the time of cytoreductive therapy were significantly more likely to develop VOD than patients with normal serum liver enzyme levels [2-4]. Other factors that have been reported to increase the risk for VOD include increasing age, certain diagnoses, estrogen-progestin therapy, amphotericin therapy during cytoreductive therapy, methotrexate therapy after transplantation, and seropositivity for cytomegalovirus [2-5, 13, 14]. In addition to increases in cytoreductive therapy dosing, several other changes in the technique of bone marrow grafting occurred in the last decade: These include the use of cyclosporine for graft-versus-host disease prophylaxis; the more extensive use of antibiotics, antifungal agents, and antiviral drugs during periods of profound immunodeficiency; and the wider use of bone marrow from unrelated and HLA-mismatched donors [15-17].

To determine the incidence of VOD after bone marrow transplantation, we prospectively followed 355 consecutive patients who underwent transplantation in Seattle, Washington. We were particularly interested in analyzing risk factors for VOD in order to explain the apparent increased incidence and severity of this complication. We also examined the relation between VOD and the failure of other organs because we have noted that multiorgan failure tends to follow the development of liver disease.

Methods

Top Methods Results Discussion Author & Article Info References

Patient Selection

We evaluated all patients referred to the Fred Hutchinson Cancer Research Center for bone marrow transplantation between August 1987 and August 1988. Three hundred and sixty patients received cytoreductive therapy in preparation for bone marrow grafting. Five patients who died before bone marrow infusion were excluded from the analysis. Thus, 355 patients were prospectively studied for the development of liver toxicity and organ failure in the early post-transplant period. Four patients received a second marrow graft within the year, but only data from the first transplant were analyzed. Underlying diseases for which transplantation was done included acute myelocytic leukemia (n = 98), acute lymphocytic leukemia (n = 61), chronic myelogenous leukemia (n = 94), lymphoma (n = 65), other malignant conditions (n = 16), and other hematologic disorders (n = 21).

Bone Marrow Transplant Techniques

The methods used at our institution have been previously reported [8, 9, 15, 18]. Briefly, patients receive chemotherapy or chemoradiation therapy before infusion of bone marrow. By convention, the day of bone marrow infusion is considered "day zero" and all post-transplant events are dated from this day. Patients who receive allogeneic marrow are given prophylactic therapy, usually with cyclosporine and methotrexate, for graft-versus-host disease [15].

Definition of Veno-occlusive Disease

A diagnosis of VOD was made according to the criterion we proposed in 1984, that is, the occurrence of two of the following events within 20 days of transplantation: hyperbilirubinemia (total serum bilirubin > 34.2 µmol/L [2 mg/dL]), hepatomegaly or right upper quadrant pain of liver origin, and sudden weight gain (> 2% of baseline body weight) because of fluid accumulation [2, 19]. No other explanation for these signs and symptoms could be present at the time of diagnosis. Patients were classified as having liver disease of uncertain cause if liver disease developed that could be explained by graft-versus-host disease, sepsis syndrome (fever and hypotension), cardiac failure, or tumor infiltration. Patients who developed both mild hyperbilirubinemia (total serum bilirubin < 34.2 µmol/L [2 mg/dL]) and weight gain below the threshold criterion of 2% were also placed in the "uncertain cause" category. Patients who died before day 5 were also classified as having disease of uncertain cause because they did not live long enough for liver disease to become clinically apparent. Patients were classified as showing no liver disease if no liver abnormalities were observed within 20 days of marrow infusion.

Clinical Course and Outcome of Veno-occlusive Disease

Patients who met criteria for the diagnosis of VOD were evaluated for severity as follows: Patients were classified as having mild disease if they showed no apparent adverse effect from liver disease; required no medications for diuresis of excessive fluid or for hepatic pain; and had completely reversible signs, symptoms, and laboratory abnormalities. Patients were classified as having moderate VOD if they had an adverse effect from liver disease; required sodium restriction and diuretics to minimize signs of fluid excess (edema, ascites, cardiopulmonary congestion) or medication to alleviate pain from hepatomegaly; and eventually showed a complete resolution of all signs of liver damage (a return of weight to baseline, a decrease in liver size, and a decrease in total serum bilirubin to < 34.2 µmol/L [2 mg/dL]). Patients were classified as having severe VOD if they showed an adverse effect from liver disease, and signs, symptoms, and laboratory values did not resolve before day 100 or the patient died, whichever occurred first. Death was not a requirement for assignment to the severe VOD category.

Risk Factors for the Development of Veno-occlusive Disease

Appendix Table 1 summarizes all risk factors analyzed for the development of VOD.

One of us reviewed referral correspondence and interviewed and examined each patient at his or her arrival to the center. Medications given before the start of cytoreductive therapy were recorded. Then, starting with day 1 of cytoreductive therapy, each patient's signs, symptoms, medications, laboratory and radiologic results, and clinical findings were noted on a daily basis until day 20 after transplantation.

We analyzed three categories of factors that could be related to the development of VOD: pretransplant factors, transplant decision factors, and clinical course factors. Pretransplant factors were those present before the start of cytoreductive therapy. Included in this category were demographic variables and variables derived from a detailed medical history, including information about individual chemotherapy agents as well as combination chemotherapy delivered before patients arrived at our center. Transplant decision factors were those decided on for each patient before the start of cytoreductive therapy. Such factors included the dose and type of cytoreductive therapy, the type of transplant, and the HLA match between donor and recipient. The dose rate of total body irradiation for most patients (271 of 277) was 6 to 7 cGy/min using opposing cobalt sources. Six patients received total body irradiation at 12 cGy/min using a linear accelerator. For patients receiving a total body irradiation dose of more than 12 Gy, we analyzed whether radiation delivered in divided daily doses (fractionated) or in multiple doses per day (hyperfractionated) had an effect on the incidence of VOD. Clinical course factors were those occurring after the start of cytoreductive therapy but before the clinical appearance of liver disease. Such factors included all medications received by each patient and the occurrence of fever.

Evaluation of Renal, Cardiac, Pulmonary and Neurologic Events

Organ dysfunction was scored on a daily basis (to day 20 after transplantation). We noted the day of onset of organ dysfunction and its severity, using the following definitions: Renal insufficiency was defined by a doubling of the baseline creatinine level (that is, the lowest serum level 1 to 3 days before bone marrow infusion). Renal failure was defined by a creatinine level of 265 µmol/L (3 mg/dL), a blood urea nitrogen level 28.6 mmol/L (80 mg/dL), or the need for hemodialysis. Cardiac failure was defined by radiographic evidence of an increased cardiac size or the development of pulmonary vascular congestion that was not present at baseline. Pleural effusions were defined by radiographic evidence in the pleural space, and pulmonary infiltrates by radiographic evidence of diffuse interstitial infiltrates that were not present at baseline. The need for oxygen support was based on documentation of hypoxemia by arterial blood gas determination or oxygen desaturation by oximetry. Mechanical ventilation was defined by the use of endotracheal intubation and ventilation to sustain life. Confusion or disorientation was defined by alterations in mental status. Bleeding requiring transfusion support was defined as nasal, oropharyngeal, intestinal, or urinary bleeding accompanied by a fall in hematocrit and requiring red blood cell support.

Platelet Transfusion Requirements

One of us recorded the quantity of platelets each patient received in transfusion on a daily basis from the start of cytoreductive therapy through day 20. Cumulative platelet transfusion requirements were calculated for each patient through day 0 and then in 5-day increments through day 20.

Relation of Veno-occlusive Disease to Death

We followed all patients with a diagnosis of VOD until death or day 100, whichever occurred first. For each patient who died, we judged whether liver dysfunction caused or contributed to the death.

Statistical Methods

Patients who did not develop liver disease had low mortality in the first 100 days; thus, only a few patients were censored from the analysis of potential VOD because of early death. Sixteen such deaths occurred, with the earliest occurring 31 days after transplantation; the median time of death was 78 days. Thus, to examine the relation between the development of VOD and baseline factors, we used contingency tables for discrete variables and the t-test for continuous variables, rather than time-to-event techniques. Where appropriate, the Fisher exact test for 2 x 2 contingency tables or Monte Carlo evaluation of the P value for a contingency table was done when more than one expected value was less than 1 [20].

Multivariate models were developed using the stepwise, step-up logistic regression model [21]. All factors related to severe VOD by univariate analysis based on a P value 0.1 were considered for entry into the multivariate model. At each stage of the multivariate model, the variable with the most additional predictive power (as measured by the smallest P value) was added as long as the P value was 0.05. Three types of variables were considered: Pretransplant variables were available before transplantation and not dependent on the therapeutic plan of the transplant team; transplant decision those factors were those determined at the time of transplantation; and clinical course variables were those factors occurring after transplantation. Models were developed in which the pretransplant variables were considered to be most fundamental; to these variables (in a stepwise fashion), the transplant decision variables were added; finally, clinical course variables were added to the model. The inclusion of so many possible risk factors in the analysis raised the "multiple comparison" problem; that is, under the null hypothesis of no relation, if many variables are examined at a 0.05 level, the probability of one or more relations being statistically significant by chance is greater than 0.05. Although the multiple comparison problem is mitigated by the dependencies in the variables considered, this remains a legitimate issue. We propose that results with P values of at least 0.002 but less than 0.05 be considered suggestive or hypothesis-generating.

The relation between the occurrence of liver disease and the failure of other organs depends on the relative timing of events. Thus, time-dependent Cox proportional-hazards regression models were used [22]. For an event that occurs at a given time (for example, classification of a patient as having VOD), the time-dependent predictive variable was 0 before the event and 1 from the time of occurrence onward.

Results

Top Methods Results Discussion Author & Article Info References

Incidence of Veno-occlusive Disease of the Liver

Of the 355 patients, 190 met our criteria for VOD (incidence rate, 54%; 95% CI, 48% to 59%). Of these 190 patients, 44 had mild disease, 92 had moderate disease, and 54 had severe disease, or 12%, 26%, and 15% of 355 patients, respectively. Eighty-six patients (24%) were classified as having "no liver disease." Seventy nine patients (22%) were classified as having liver disease of "uncertain cause." The most common reason for assigning patients to the "uncertain cause" category was the development of another liver disease within 20 days of bone marrow infusion: These other liver diseases included acute graft-versus-host disease (17 patients), sepsis syndrome (22 patients), cardiac failure (6 patients), persistent tumor infiltration of the liver (3 patients), or combinations of these processes. Twenty-eight patients were labeled as having liver disease of uncertain cause because they had two or more signs of liver dysfunction yet failed to meet our criteria for VOD. Three patients were placed in the "uncertain cause" category because they died before liver disease could become manifest.

Analysis of Risk Factors for Severe Veno-occlusive Disease

Appendix Table 2 summarizes the results of a univariate analysis of risk factors for the development of severe VOD.

Pretransplant Factors

Elevated serum transaminase levels were highly correlated with subsequent severe VOD (relative risk, 2.8 for patients with increased aspartate aminotransferase levels relative to those with normal levels). Patients with greater transaminase elevations had a higher incidence of severe VOD than those with minimal elevations. Statistical significance was marginal for a relation between severe VOD and a remote history of hepatitis, drug-related liver injury, or abdominal radiation therapy. The relation between previous marrow transplantation and severe liver toxicity was not statistically significant, even when data from the 4 patients who received two transplantations in the year of our study were added to data from 18 patients who had received previous transplantations. However, the use of either of two antimicrobial agents, vancomycin and acyclovir, in the weeks before the start of cytoreductive therapy was correlated with severe VOD (relative risk, 2.1 and 3.0, respectively).

Transplant Decision Factors

Certain cytoreductive therapy regimens were associated with a higher incidence of severe VOD. Our "standard" regimen of cyclophosphamide plus 12 Gy of fractionated total body irradiation was associated with an 8% incidence of severe liver toxicity. Regimens associated with a higher incidence included those incorporating higher doses of total body irradiation (23%); busulfan plus cyclophosphamide (32%); and BCNU (1,3'-bis [2-chloroethyl]-1-nitrosourea) cyclophosphamide, and etoposide (33%). We could not discern an effect of the radiation delivery schedule in patients receiving more than 12 Gy of total body irradiation, because the incidence of severe liver toxicity was the same in patients on fractionated or hyperfractionated schedules of irradiation at a dose rate of 6 to 7 cGy/min. Of six patients who received total body irradiation at a dose rate of 12 cGy/min, three developed liver toxicity of mild (two patients) or moderate (one patient) severity. The incidence of severe VOD was not different in autologous and allogeneic bone marrow transplant recipients (there was one syngeneic transplant recipient). However, among allogeneic bone marrow transplant recipients, patients whose marrow donor was not HLA-identical had a higher incidence of severe VOD.

Clinical Course Factors

Use of either vancomycin or amphotericin (but none of six other types of antimicrobial agents) during the administration of cytoreductive therapy was significantly related to severe VOD (relative risk, 2.6 and 2.0, respectively). At our center, vancomycin and amphotericin are most commonly prescribed for persistent fever unresponsive to broad-spectrum antibiotics. We examined the relation of fever to later liver toxicity by calculating the ratio of the number of days on which body temperature exceeded 38 °C to the total number of days from the start of cytoreductive therapy through day 5. This ratio was 0.13 ± 0.02 for patients without VOD and 0.22 ± 0.02 for patients with VOD (P < 0.001). Patients with mild or moderate VOD had a fever ratio of 0.18 ± 0.02, and those with severe VOD had a ratio of 0.32 ± 0.04. The proportion of febrile days was significantly higher for patients who developed severe VOD compared with those who did not (P < 0.001).

We sought to determine why vancomycin use and severe VOD were related, the issue being whether vancomycin had a direct hepatotoxic effect or whether the drug acts as a marker for persistent fever. We carried out a stepwise logistic regression analysis of the effects of the following factors on the incidence of severe VOD: fever ratio, total vancomycin dose per kg of body weight, and the number of days on which vancomycin was given. In the stepwise model, only the fever ratio was related to severe VOD (relative risk, 15.8; CI, 4.4 to 57.2), suggesting that the dose and duration of vancomycin therapy were not important contributors to severe VOD. This analysis supports the hypothesis that vancomycin therapy is predictive of severe VOD because it is a marker for persistent fever.

Therapy with granulocyte-macrophage colony-stimulating factor (GM-CSF) appeared to be protective against the development of severe VOD; none of 21 patients who received GM-CSF developed this complication, but 21% of patients who did not receive GM-CSF did. However, selection bias may explain this result because only patients with normal pretransplant serum liver enzyme levels received GM-CSF.

Stepwise Logistic Regression (Multivariate) Analysis

Results of a logistic regression analysis of risk factors for severe VOD are shown in Table 1. To assess the effect of disease status (relapse or remission) in patients with leukemia as an independent predictive variable, we carried out a preliminary stepwise regression analysis of pretransplant risk factors for severe VOD in 201 patients with a diagnosis of leukemia. Pretransplant factors considered for entry in the model included disease status (relapse or remission), an indicator variable for the diagnosis of chronic myelogenous leukemia, the interaction between the diagnosis of chronic myelogenous leukemia and leukemia phase (chronic phase or blast crisis), elevated serum transaminase levels, transfusion history, previous radiation therapy to the abdomen, past history of hepatitis, previous drug-related liver injury, and vancomycin or acyclovir therapy. Only elevated serum transaminase levels, vancomycin use, and acyclovir use were significantly associated with severe VOD. Because leukemic relapse, diagnosis of chronic myelogenous leukemia, and phase of chronic myelogenous leukemia were not significant independent variables, subsequent analyses were done using all patients except those in the "uncertain cause" category. This sample included 276 patients.

The final stepwise logistic regression analysis was done in three phases, first considering pretransplant factors, then transplant decision factors, and then clinical course factors. Pretransplant factors considered for the model were elevated serum transaminase levels, a history of hepatitis or a positive serologic test result for a hepatitis virus, transfusion history, previous drug-related liver injury, previous radiation therapy to the abdomen, and therapy with vancomycin or acyclovir. In the next phase of the analysis, the transplant decision variables "high-dose cytoreductive therapy" and "mismatched or unrelated donor marrow" (Appendix Table 2 were considered for entry in the model. In the final phase of the analysis, clinical course variables, including vancomycin therapy, amphotericin therapy, antithymocyte globulin therapy, and GM-CSF use, were considered for entry. The independent factors predictive of severe VOD are listed in Table 1, along with their respective relative risks and 95% CIs.

The initial signs of liver toxicity were sudden weight gain and the development of hepatomegaly or liver tenderness. The median time of occurrence for both signs was day 0 (first to third quartiles, day –3 to day 2 and day –6 to day 6, respectively). Acute abdominal pain was usually perceived in the epigastrium, but that the pain originated in the liver was readily apparent on palpation of the abdomen. Other median times of occurrence were as follows: hyperbilirubinemia (total serum bilirubin, 34.2 µmol/L [2 mg/dL]), day 6 (first to third quartiles, day 2 to day 9); peripheral edema, day 7 (first to third quartiles, day 4 to 12); and ascites, day 12 (first to third quartiles, day 7 to 17). Time of onset of hyperbilirubinemia was similar for patients on regimens including total body irradiation and patients on chemotherapy-only regimens, whether time of onset was measured from the start of cytoreductive therapy or from day 0. Patients with severe VOD differed from those with reversible VOD in the amount of weight gained, time of onset of jaundice, the maximum total serum bilirubin level, and a greater frequency of edema and ascites (Table 2).

Platelet Transfusion Requirements

Platelet transfusion requirements were significantly higher among patients with VOD than among patients without liver disease. The increased requirement was apparent by day 0, even before the onset of clinical liver disease (12.3 ± 16.9 compared with 5.4 ± 9.3 cumulative platelet units; P = 0.0001). The most striking differences were in patients destined to develop severe VOD, (mean transfusion requirement through day 0, 21.7 ± 22.2 platelet units). Differences in cumulative platelet requirements persisted through day 20.

Relation between Liver Toxicity and Multiorgan Failure

Development of VOD was significantly related to every variable related to organ failure that we measured (Table 3). Particularly striking was the high prevalence of organ failure among patients with VOD and, conversely, the near absence of renal and respiratory failure among patients without liver disease. Each of these end points of organ failure was more frequent among patients with severe VOD than among patients with mild or moderate disease (see Table 3).

The time of onset of liver disease and of the failure of specific organs is summarized in Table 4. These values are averages, but variation occurred from patient to patient as shown by the standard deviations. To test the hypothesis that signs and symptoms of liver disease predicted the failure of other organs, the occurrence of these signs and symptoms were treated as time-dependent variables in the Cox proportional-hazards model. The occurrence of a bilirubin level above 34.2 µmol/L (2 mg/dL) and the occurrence of a weight gain of more than 2% were both statistically significant predictors of the onset of pleural effusions, cardiac failure, pulmonary infiltrates and confusion (P < 0.001 for each analysis; data not shown). The occurrence of a bilirubin level above 34.2 µmol/L was a statistically significant predictor of renal insufficiency (relative risk, 4.9), renal failure (relative risk, 43.4), and bleeding requiring transfusions (relative risk, 2.3). The frequency with which liver dysfunction preceded the failure of other organs strongly supported the hypothesis that liver disease was causally related to organ failure (data not shown).

Clinical Course and Outcome of Patients with Veno-occlusive Disease

For patients with VOD, the mortality rate before day 100 was 39% (78 of 190 patients): 98% (53 of 54 patients) among patients with severe VOD, 23% (21 of 92 patients) among patients with moderate VOD, and 9% (4 of 44 patients) among patients with mild VOD. For comparison, mortality before day 100 for patients who did not develop liver disease was 19% (16 of 86 patients). Mortality before day 100 in patients categorized as having liver disease of uncertain cause was 56% (44 of 79 patients). Of 190 patients who developed VOD, 78 died before day 100. We judged that liver dysfunction caused or contributed to the deaths of 54 of these patients and that liver dysfunction played no role in the deaths of 24 patients.

Discussion

Top Methods Results Discussion Author & Article Info References

In our series, the incidence of liver disease in the 20 days after transplantation was high. We classified 190 of 355 patients (54%) as having VOD, a clinical syndrome characterized by hepatomegaly, weight gain, and jaundice. The incidence of VOD has increased markedly since 1980, when we reported a rate of 21% based on the same criteria used in the current study [2]. The increase in incidence from 21% to 54% may be explained by the regimens of cytoreductive therapy that are currently used to prepare patients for marrow grafting at our center. During the year 1979-1980, 22% of 253 persons were on one of the following three regimens: cyclophosphamide plus total body irradiation over 12 Gy; cyclophosphamide plus busulfan; or BCNU, cyclophosphamide, and etoposide [2]. In our current study, these regimens were prescribed to 58% of 355 patients to reduce the rate of recurrent malignancy.

Severe VOD occurred in 15% of our patients. All but one such patient died before day 100. Multivariate analysis of risk factors confirmed that high-dose cytoreductive therapy (relative risk, 2.8) and pretransplant transaminase elevation (relative risk, 4.6) were independent predictors of severe VOD. Ours is the fourth largest prospective study to find an association between pretransplant transaminase elevations and subsequent development of VOD [2-4]. In our series, the risk for severe VOD was higher among patients who had minor serum transaminase elevations compared with patients who had normal serum transaminase levels. The risk for severe VOD increased as transaminase levels increased.

Why should hepatitis make the liver more susceptible to VOD? Our first hypothesis was that drug metabolism was abnormal in patients with even mild chronic hepatitis, resulting either in high levels of cytotoxic metabolites of chemotherapy drugs in zone 3 of the liver acinus or in failure to detoxify these metabolites [7]. Substantial evidence exists that blood levels after identical oral doses of busulfan vary widely among patients and that toxicity is related to abnormally high blood levels [23, 24]. Recent studies have found that the pharmacokinetics of busulfan vary with age, children having lower steady-state concentrations and higher clearance rates than adults [23, 25-27]. It is not clear whether this variability in blood levels is related to intestinal absorption or to altered hepatic metabolism of busulfan. Hepatitis may affect the metabolism of drugs used to prepare patients for transplantation, because both the glutathione S-transferases (involved in the detoxification of free oxygen radicals) and cytochrome P450 mixed-function oxygenases (which bioactivate cyclophosphamide) are predominantly found in hepatocytes in zone 3 of the liver acinus [28-30]. Depletion of hepatic and endothelial glutathione may be related to the development of VOD: Busulfan, total body irradiation, and BCNU (components of the three most hepatotoxic regimens in our study) affect glutathione levels [31, 32], and, in an animal model of VOD, glutathione depletion enhanced the hepatic toxicity of cytotoxic drugs [33]. The hypothesis of altered drug metabolism is attractive from a prevention perspective because doses of cytotoxic drugs can be adjusted for an individual patient's metabolism for that drug, thereby lessening liver toxicity [27]. Manipulation of the glutathione and cytochrome P450 enzyme systems to lessen hepatic toxicity in bone marrow transplant recipients may also be possible, but physicians contemplating the use of these maneuvers must consider the problem of tumor relapse.

An alternative but not mutually exclusive hypothesis to explain the relation between pretransplant transaminase elevation and VOD states that patients with hepatitis have abnormal sinusoidal and venular endothelium that is more susceptible to chemoradiation injury than normal endothelium. The hepatic endothelium is approximately three times more sensitive to a single dose of radiation than are hepatocytes [34, 35]. Recent studies have shown that hepatic endothelium in patients with hepatitis expresses adhesion molecules [36, 37] and procoagulant tissue factor [38, 39], effects mediated by inflammatory cytokines [37, 40, 41]. Patients with chronic hepatitis also have increased expression of mRNA encoding for transforming growth factor-ß, a promoter of collagen synthesis by Ito cells [42]. Because the primary site of injury after cytoreductive therapy is the terminal hepatic venule and because endothelial cells are more sensitive to the effects of irradiation than hepatocytes, we theorize that activated hepatic endothelium in patients with hepatitis is more readily damaged by cytoreductive therapy. This "abnormal endothelium" hypothesis is attractive in that pharmacologic and biologic agents may be capable of modulating cytokine effects on hepatic endothelium without affecting the antitumor effects of cytoreductive therapy.

Some investigators have not found an association between pretransplant transaminase elevations and VOD in bone marrow transplant recipients [14, 43-47]. Patients with thalassemia, for example, have a high prevalence of chronic hepatitis because of exposure to hepatitis viruses in transfusion products, but they do not have a high incidence of VOD after bone marrow grafting [44, 46]. It may be that variability in the absorption and metabolism of oral busulfan in children [25-27] results in fewer toxic metabolites within the sinusoids of the liver. Lower-dose cytoreductive therapy is the likely explanation for a low incidence of liver toxicity in other series (cyclophosphamide alone [43] or total body irradiation given at a slow dose rate (2 cGy/min [44]). Our series showed that fractionated and hyperfractionated radiation schedules yield an identical incidence of severe VOD, but we could not examine irradiation dose rate as a risk factor because most patients received total body irradiation at a dose of 6 to 7 cGy/min.

Another recent study has shown a low incidence of VOD (with no severe cases) in patients receiving T-cell-depleted allogeneic transplantation after conditioning therapy with cyclophosphamide and 12 to 14 Gy of fractionated total body irradiation [14]. In that study, the possibly protective effect of T-cell depletion on the development of VOD could not be separated from the fact that these patients were not exposed to cyclosporine or methotrexate, which are both potential hepatotoxins. Our study showed that mismatched donor marrow was a risk factor for severe VOD, lending credence to the hypothesis that infusion of alloimmune T cells may lead to greater cytokine release, contributing to liver toxicity. The possibly hepatotoxic effects of cyclosporine and methotrexate were not risk factors for severe VOD in our study, but other studies have shown that these drugs may cause liver dysfunction after transplantation [13, 15]. The clinical onset of VOD was apparent in many of our patients before either bone marrow infusion or cyclosporine-methotrexate dosing, suggesting that events that occur during cytoreductive therapy are the most important in the pathogenesis of VOD. Another explanation for the discrepancy in incidence rates for liver toxicity may be that persistent fever during cytoreductive therapy (which was predictive of severe VOD) may not have been accounted for in these studies. One previous study of patients who developed VOD after conditioning therapy with busulfan plus cyclophosphamide noted an association between amphotericin use and severe VOD [5].

We found an unexpected association between severe VOD and the use of several antimicrobial agents before and after the start of cytoreductive therapy. Only pretransplant acyclovir use (relative risk, 4.8) and vancomycin use during cytoreductive therapy (relative risk, 2.9) proved to be independent risk factors by multivariate analysis. The association between these drugs and severe liver toxicity has two possible explanations: The drugs could be hepatotoxic themselves, or they could be markers for an inflammatory process that affects the liver. At our institution, vancomycin is given most frequently on an empiric basis for fever that is unresponsive to broad-spectrum antibiotics. Vancomycin has been reported to cause only mild elevations in serum transaminase levels when given as empiric therapy for fever in granulocytopenic patients [48, 49]. We found that fever was highly correlated with VOD; that is, patients who developed VOD spent more of the days preceding the onset of liver disease in a febrile state than did patients who did not develop VOD. When we examined the interplay among fever, vancomycin dose, and duration of vancomycin therapy, only fever was a statistically significant predictor of severe VOD. This finding supports the hypothesis that vancomycin use is a marker for a systemic inflammatory process that affects the liver. We speculate that acyclovir use is also a marker for the inflammatory response caused by herpesvirus infections.

Recent studies have shown an association between the presence of inflammatory cytokines (especially tumor necrosis factor-) in blood samples and the toxic effects of cytoreductive therapy [50]. Inhibition of tumor necrosis factor release with the drug pentoxifylline has also been reported to alleviate toxicity in bone marrow transplant recipients [51]. A role for tumor necrosis factor in the pathogenesis of VOD is consistent with several known effects of the factor, notably the down-regulation of thrombomodulin, prostaglandins E₁ and E₂, and protein S, and the up-regulation of cellular adhesion molecules and platelet-derived growth factor [38, 52, 53]. Because evidence exists that an intrahepatic prothrombotic state contributes to the obstruction of sinusoidal blood flow in VOD [6, 54-57], the procoagulant effects of tumor necrosis factor are of particular interest. These include effects on protein C (through decreases in thrombin-thrombomodulin complex), increased local synthesis of tissue factor, and initiation of the coagulation cascade [38, 53, 58]. Bacterial endotoxins and inflammatory cytokines may also directly affect hepatocytes and bile ductules [59, 60]. The converse may also apply, because the liver is both a major source of and a clearance organ for circulating cytokines [60].

Multiorgan failure was significantly more common among patients with VOD than among patients without VOD. Over half of the patients with severe VOD developed renal failure, cardiac failure, and pleural effusions. Pulmonary failure, evinced by the need for mechanical ventilation, occurred in 43% of patients with severe VOD. The high mortality rate for patients with severe VOD is not surprising given the high prevalence of cardiopulmonary and renal failure in this population. We judged that liver disease contributed to the deaths of 47 of 53 patients with severe VOD who died. In 16 of these patients, VOD was the only apparent cause of liver dysfunction; however, 37 patients developed other liver diseases after the onset of VOD. The most common of these other causes of liver disease was graft-versus-host disease. Although graft-versus-host disease affects 30% to 70% of recipients of allogeneic bone marrow, the prevalence of severe graft-versus-host disease in allogeneic recipients with liver and renal toxicity is higher because liver and renal dysfunction preclude the use of full-dose cyclosporine and methotrexate for prophylaxis of graft-versus-host disease [61].

Our clinical impression that liver disease usually develops before the failure of the kidneys, heart, and lungs is supported by the results of our statistical analysis, which show that jaundice and sudden weight gain are highly likely to precede the failure of these organs. This chronologic pattern suggests a causal relation between liver dysfunction and organ failure. Liver toxicity has a profound effect on renal sodium excretion in these patients, as shown by a low fractional excretion of sodium coincident with rapid gains of weight at day 1, on average. After an interval of 5 to 7 days, patients show diffuse pulmonary infiltrates on chest radiograph, pleural effusions, peripheral or presacral edema, and radiologic signs of heart failure. After another 5 to 7 days, hypoxemia, renal insufficiency, ascites, and respiratory failure develop. A retrospective review found that both hyperbilirubinemia and amphotericin use were independent risk factors for renal failure after bone marrow transplantation [62]. Other series have also shown a high incidence of multiorgan failure in bone marrow transplant patients, as well as evidence that pulmonary dysfunction and pleural effusions are related to VOD [11, 12, 47, 63]. A central role for liver dysfunction in multiorgan failure has also been reported in patients who are not bone marrow transplant recipients [64-68].

The mechanisms whereby liver dysfunction might lead to multiorgan failure have been reviewed recently [68, 69]. Failure of the hepatic reticuloendothelial system is thought to be important; this macrophage system normally handles enteric bacteria and endotoxins [59]. Tissue macrophages (Kupffer cells) are destroyed by cytoreductive therapy and are later replaced by cells of marrow donor origin [70]. Liver dysfunction appears to have a direct effect on intestinal integrity, allowing movement of bacterial products into the portal circulation where they initiate an inflammatory cascade [71]. The liver also produces opsonizing proteins, which are important in the clearance of bacteria, and anticoagulant proteins (proteins C and S and antithrombin III), which prevent in-situ thromboses. Liver dysfunction, in turn, may lead to the deficient clearance of bacteria, bacterial products, and cytokines and to a procoagulant state, both of which have been invoked to explain some of the manifestations of multiorgan failure [68, 69]. Our experience in marrow transplant recipients suggests that when failure of liver host defenses are combined with hepatorenal and hepatopulmonary syndromes [72], the case fatality rate is high. In contrast to other settings in which multiorgan failure develops, such as sepsis and trauma, the role of neutrophils is negligible in the bone marrow transplant setting, because cytoreductive therapy results in agranulocytosis during the time that organ failure is developing.

The finding that platelet transfusion requirements were higher among patients with VOD than among those without VOD confirms a previous study [73]. In contrast, a large prospective study of bone marrow transplant recipients failed to find this relation [3]. Our data show a twofold increase in the platelet requirements, even before the day of marrow infusion, among patients who later showed clinical evidence of VOD compared with patients without VOD. Immunohistologic studies have found no evidence of platelet antigens enmeshed in the occlusive lesions within the liver of bone marrow transplant recipients with VOD [6]. We speculate that obstruction of liver blood flow leads to an increased destruction of platelets in the spleen.

In summary, our prospective study has documented that over half of our marrow transplant recipients developed veno-occlusive disease of the liver. Severe VOD developed in 15% of cases and was associated with multiorgan failure and death. Patients at high risk for this complication can be recognized before transplantation, because elevated serum transaminase enzyme levels and certain high-dose regimens are predictive of severe VOD. Infection and persistent fever before and during cytoreductive therapy are additional independent risk factors for subsequent severe VOD. The challenge for the future will be to devise strategies to protect the liver from the toxic effects of high-dose cytoreductive therapy while preserving antitumor effects.

Abbreviation

VOD = veno-occlusive disease (of the liver)