Effectiveness of antibacterial prophylaxis during induction chemotherapy in children with acute lymphoblastic leukemia.

BACKGROUND: Pediatric patients receiving induction chemotherapy for newly diagnosed acute lymphoblastic leukemia (ALL) are at high risk of developing life-threatening infections. We investigated whether uniform antibacterial guidelines, including mandatory antibacterial prophylaxis in afebrile patients during induction, decreases the incidence of microbiologically documented bacteremia. METHODS: Between 2012 and 2015, 230 patients with newly diagnosed ALL (aged 1-21) were enrolled on Dana-Farber Cancer Institute ALL Consortium Protocol 11-001 (DFCI 11-001). Induction therapy, regardless of risk group, included vincristine, prednisone, doxorubicin, methotrexate, and PEG-asparaginase. Afebrile patients received fluoroquinolone prophylaxis at the initiation of induction and those presenting with fever received broad-spectrum antibiotics; antibiotics were continued until blood count recovery. Rates of documented bacteremias and fungal infections on DFCI 11-001 were compared to those on the predecessor protocol (DFCI 05-001), which included the same induction phase without antibiotic prophylaxis guidelines. RESULTS: Sixty-six (28.7%) patients received fluoroquinolone prophylaxis, the remaining patients received broad-spectrum antibiotics. Twenty-four (36.4%) patients on prophylaxis developed fever and seven (10.6%) developed bacteremia. The overall rate of infection during induction on DFCI 11-001 was lower than on DFCl 05-001 (14.3% vs. 26.3%, P < 0.0001) due to a decreased rate of bacteremia (10.9% vs. 24.4%, P < 0.0001). The rate of fungal infections (4.8% vs. 3.6%) and induction death (0.9% vs. 2%) was not significantly different. CONCLUSION: For children with newly diagnosed ALL, uniform antibiotic administration until blood count recovery, including fluoroquinolone prophylaxis for afebrile patients, reduced the incidence of bacteremia during the induction phase. Larger, randomized studies should be performed to confirm these findings.

Polymorphisms of ABCC5 and NOS3 genes influence doxorubicin cardiotoxicity in survivors of childhood acute lymphoblastic leukemia.

Anthracyclines are efficient chemotherapy agents. However, their use is limited by anthracycline-induced cardiotoxicity (CT). We investigated the influence of polymorphisms in doxorubicin metabolic and functional pathways on late-onset CT as estimated by echocardiography in 251 childhood acute lymphoblastic leukemia (cALL) patients. Association analyses revealed a modulating effect of two variants: A-1629 T in ABCC5, an ATP-binding cassette transporter, and G894T in the NOS3 endothelial nitric oxide synthase gene. Individuals with the ABCC5 TT-1629 genotype had an average of 8-12% reduction of ejection (EF) and shortening fractions (SF; EF: P<0.0001, and SF: P=0.001, respectively). A protective effect of the NOS3 TT894 genotype on EF was seen in high-risk patients (P=0.02), especially in those who did not receive dexrazoxane (P=0.002). Analysis of an additional cohort of 44 cALL patients replicated the ABCC5 association but was underpowered for NOS3. In summary, we identified two biomarkers that may contribute to cALL anthracycline CT risk stratification.

Role of CD81 and CD58 in minimal residual disease detection in pediatric B lymphoblastic leukemia.

INTRODUCTION: Minimal residual disease (MRD) in B lymphoblastic leukemia has been demonstrated to be a powerful predictor of clinical outcome in numerous studies in both children and adults. In this study, we evaluated 86 pediatric patients with both diagnostic and remission flow cytometry studies and compared expression of CD81, CD58, CD19, CD34, CD20, and CD38 in the detection of MRD. METHODS: We evaluated 86 patients with B lymphoblastic leukemia who had both diagnostic studies and remission studies for the presence of MRD using multicolor flow cytometry. We established our detection limit for identifying abnormal lymphoblasts using serial dilutions. We also compared flow cytometry findings with molecular MRD detection in a subset of patients. RESULTS: We found that we can resolve differences between hematogones and lymphoblasts in 85 of 86 cases using a combination of CD45, CD19, CD34, CD10, CD20, CD38, CD58, and CD81. Our detection limit using flow cytometry is 0.002% for detecting a population of abnormal B lymphoblasts. Comparison with MRD assessment by molecular methods showed a high concordance rate with flow cytometry findings. CONCLUSIONS: Our study highlights importance of using multiple markers to detect MRD in B lymphoblastic leukemia. Our findings indicate that including both CD58 and CD81 markers in addition to CD19, CD34, CD20, CD38, and CD10 are helpful in MRD detection by flow cytometry.

Subset derivation of T-cell acute lymphoblastic leukemia in man.

Normal human peripheral blood T cells can be characterized as belonging to either the TH+2 or TH-2 T-cell subset. Approximately 20% of T cells are TH+2, whereas 80% are TH-2 utilizing specific heteroantisera. To determine shether human T-cell acute lymphoblastic leukemia (T-ALL) cells belong to one or another T-cell subset, cell surface phenotyping was performed on tumor populations from 25 patients with T-ALL. Tmuor cells from these 25 individuals were either TH+1 or TH-2, but not both. 5 of 25 patients had TH+2 T-ALL cells. These TH+2 tumor populations were found exclusively in children and often without an accompanying thymic mass. TH-2 T-ALL, in contrast, occurred in both children and adults and was almost always associated with thymic enlargement. Although children with TH+2 T-ALL had as high or higher peripheral blast counts on presentation than their TH-2 T-ALL counterparts, overall survival was greater for the TH+2 group (greater than 36 mo) than the TH-2 group (less than 12 mo). These studies demonstrate that T-cell leukemias in man arise from distinct T-cell subsets and that cell surface characterization of T-cell malignancies may provide useful clinical data related to prognosis.

Induction of human B cell antigens in non-T cell acute lymphoblastic leukemia.

Leukemic cells from 70% of patients with Ia+CALLA+ non-T cell acute lymphoblastic leukemia (ALL) express an antigen (B1) found on all normal B lymphocytes. In this study, ALL cells that do not express the B1 antigen were studied in an attempt to further elucidate the cellular lineage of these tumors. Non-T cell ALL lines and tumor cells isolated from patients with non-T cell ALL that are Ia + CALLA + B1- were studied in vitro with a variety of agents known to promote cellular differentiation. Phorbol diester (TPA) or phytohemagglutinin conditioned leukocyte culture media were capable of inducing the expression of B1 on all four non-T cell ALL lines tested. In contrast, B1 could not be induced under the identical conditions on a promyelocytic leukemia line or a T cell lymphoblastic leukemia line. With the induction of B1 on non-T cell ALL lines, cytoplasmic mu-heavy chain (c mu) became undetectable, whereas the expression of CALLA and Ia were unchanged. The expression of B1 was accompanied by a decrease of cellular proliferation and DNA synthesis, but not significant morphologic changes were noted. In addition, no other B or T cell antigens were detected. The cellular origin of non-T cell ALL was further investigated using tumor cells isolated from leukemic patients. Tumor cells from eight patients with Ia + CALLA + B1-c mu- ALL could be induced in vitro with TPA to express both B1 and c mu. In contrast, cells from five patients with Ia + CALLA-B1-c mu- non-T cell ALL could not be induced with TPA to express CALLA, B1 or c mu. These studies suggest that the non-T cell ALL are heterogeneous and represent a spectrum of early B cell differentiation including the pre- pre-B cell (Ia + CALLA + B1-c mu-), the intermediate pre-B cell (Ia + CALLA +B1 + c mu-), and finally the "true" pre-B cell (Ia + CALLA + B1 + c mu+). The cellular origin of the remaining Ia + CALLA-B1-c mu- form of non-T cell ALL (20%) is still unknown.

B cell origin of non-T cell acute lymphoblastic leukemia. A model for discrete stages of neoplastic and normal pre-B cell differentiation.

The expression of B cell associated and restricted antigens on tumor cells isolated from 138 patients with non-T cell acute lymphoblastic leukemia (non-T cell ALL) was investigated by flow cytometric analysis by means of a panel of monoclonal antibodies. Tumor cells from these patients could be assigned to one of four subgroups: human leukocyte antigen-DR-related Ia-like antigens (Ia) alone (4%, stage I); IaB4 (14%, stage II); IaB4CALLA (33%, stage III); and IaB4CALLAB1 (49%, stage IV). The expression of B cell-restricted antigens (B4 and B1) and rearrangements of Ig heavy chain genes provided strong evidence for the B cell lineage of stages II, III, and IV tumors. The lineage of the Ia alone group is still unknown. The B4 antigen was expressed on approximately 95% of all non-T cell ALLs tested, and given its absence on T cell and myeloid tumors, it appears to be an exceptional marker to define cells of B lineage. The demonstration that Ia alone, IaB4, IaB4CALLA, and IaB4CALLAB1 positive cells can be readily identified by dual fluorescence analysis in normal fetal and adult bone marrow provided critical support for the view that these leukemic pre-B cell phenotypes were representative of the stages of normal pre-B cell differentiation. It was interesting that the IaB4+ cell was more frequently identified in fetal bone marrow than in adult marrow, whereas the predominant cell found in adult marrow expressed the IaB4CALLAB1 phenotype. These data suggest that the leukemogenic event may be random, since the predominant pre-B cell leukemic phenotype appears to correspond to the normal pre-B cell phenotype present in these hematopoietic organs. Our observations provide an additional distinction between adult and childhood ALL, since these studies show that most non-T cell ALLs seen in children less than 2 yr old are of stage II phenotype, whereas the majority of non-T ALLs in adults are of stage IV phenotype. Finally, it should be noted that the present study suggests that the analysis of leukemic B cell phenotypes and their normal counterparts can provide a mechanism for the investigation and orderly definition of stages of pre-B cell differentiation in man.

Silencing of the tumor suppressor gene FHIT is highly characteristic for MLL gene rearranged infant acute lymphoblastic leukemia.

MLL rearranged acute lymphoblastic leukemia (MLL) is an aggressive type of acute lymphoblastic leukemia (ALL), diagnosed predominantly in infants (<1 years of age). Since current chemotherapy fails in >50% of patients with MLL, new therapeutic strategies are desperately needed. For this, understanding the biological features characterizing MLL is necessary. Analysis of gene expression profiles revealed that the expression of the tumor suppressor gene FHIT is reduced in children with MLL rearranged ALL as compared to ALL patients carrying germ line MLL. This finding was confirmed by quantitative real-time PCR. In 100% of the infant MLL cases tested, methylation of the FHIT 5'CpG region was observed, resulting in strongly reduced mRNA and protein expression. In contrast, FHIT methylation in infant and non-infant ALL patients carrying germ line MLL was found in only approximately 60% (P< or =0.004). FHIT expression was restored upon exposing leukemic cells to the demethylating agent decitabine, which induced apoptosis. Likewise and more specifically, leukemic cell death was induced by transfecting MLL rearranged leukemic cells with expression vectors encoding wild-type FHIT, confirming tumor suppressor activity of this gene. These observations imply that suppression of FHIT may be required for the development of MLL, and provide new insights into leukemogenesis and therapeutic possibilities for MLL.

Second neoplasms after Wilms' tumor in childhood.

The frequency of second neoplasms was examined among 487 Wilms' tumor patients treated at the Dana-Farber Cancer Institute and Children's Hospital of Boston, 1927-81. Thirty study patients (6%) developed second primary tumors: 11 cancers, 16 benign tumors, and 3 borderline neoplasms. Cumulative probability of a second cancer was 18% (standard error, 6%) in 34 years after diagnosis of Wilms' tumor. The subgroup of 412 patients who had received radiotherapy for Wilms' tumor developed all 11 second cancers, which included 1 skin carcinoma, 1 acute leukemia, and 9 solid internal cancers (expected, 0.7 cancers other than skin carcinoma; P less than .001). After exclusion of the secondary leukemia, all but 1 second cancer arose within the radiotherapy field. Concurrent therapy with dactinomycin did not reduce the risk of a radiation-associated cancer. Second cancer was the cause of death in 7 patients.

Dynamic parameters of membrane lipids in normal and leukemic human lymphocytes isolated from peripheral blood and bone marrow.

The degree of microviscosity (eta), and lipid fluidity (LFU) of cellular membranes of normal and leukemic lymphocytes obtained from peripheral blood and bone marrow of normal donors and acute lymphatic leukemic (ALL) patients was quantitatively monitored by fluorescence polarization analysis with the aid of the fluorescent lipophilic probe 1,6-diphenyl-1,3,5-hexatriene when embedded in cellular membranes of intact cells. The results have shown a marked decrease in eta and a significant increase in LFU in lymphocytes obtained from both peripheral blood and bone marrow of ALL patients at admission when compared to both T- and B-lymphocytes obtained from peripheral blood of normal donors. Moreover, both dynamic parameters, eta and LFU, show normal characteristic values in lymphocytes obtained from bone marrow of ALL patients in complete hematological remission. Since in few cases a decrease in eta and an increase in LFU were observed in bone marrow lymphocytes isolated from ALL patients in remission, the possibility that these dynamic parameters may serve as a diagnostic tool for an early detection of a new relapse is discussed.

Clinical utility of leukemia cell terminal transferase measurements.

Interest in the DNA-synthetic enzyme terminal deoxynucleotidyl transferase (TdT) has developed from two sets of observations: first, in normal animals, it occurs only in immature thymic lymphocytes and in a subpopulation of bone marrow lymphocytes; second, it is present in the blast cells of almost all patients with acute lymphoblastic leukemia. A prospective trial to evaluate blast cell TdT as a predictor of responsiveness to vincristine and prednisone in 30 Philadelphia chromosome-positive patients with blastic chronic myelogenous leukemia was undertaken. Eleven of 16 TdT-positive patients responded, whereas only one of 14 TdT-negative patients showed improvement. Among TdT-positive patients under the age of 50 years, the response rate was 78%. Enzyme-negative patients under the age of 50 had an 11% response rate. Blast cell morphology (i.e., lymphoblastic versus myeloblastic) had no significant correlation with either responsiveness or TdT activity. These results suggest that blast cell TdT activity may identify leukemic patients who are likely to respond to vincristine and prednisone irrespective of their conventional classification.

Influence of intensive asparaginase in the treatment of childhood non-T-cell acute lymphoblastic leukemia.

Between June 1977 and December 1979, 72 evaluable patients with childhood non-T-cell acute lymphoblastic leukemia were induced into complete remission using vincristine, prednisone, and doxorubicin. All received asparaginase consolidation and central nervous system prophylaxis with cranial irradiation and intrathecal methotrexate. All patients then received prolonged intensification with vincristine, prednisone, and doxorubicin, and half of them were randomized to receive weekly high-dose asparaginase. Continuation therapy was with vincristine, prednisone, methotrexate, and 6-mercaptopurine. After a median follow-up of 57 months, there were four remission deaths and 25 relapses. Central nervous system relapse was the first event in 4% of patients. There were fewer treatment failures in the asparaginase-treated group [2-sided, p = 0.04 (0.07 controlling for standard and high-risk groups)]. Asparaginase toxicity occurred in six patients (8%) and was self-limited, but it precluded further use of the drug in those patients. The major toxicity of this treatment program was drug-induced cardiomyopathy which occurred in 10 patients (14%) and was fatal in three of them. In summary, we conclude that the intensive use of high-dose asparaginase has an important role in the treatment of children with acute lymphoblastic leukemia. The morbidity of multiple doses of doxorubicin outweighed its antileukemic advantage in standard-risk patients.

Use of monoclonal antibodies as diagnostic and therapeutic reagents in acute lymphoblastic leukemia.

A monoclonal antibody specific for a common acute lymphoblastic leukemia (ALL) antigen has been generated and characterized. This antibody (J-5) is reactive with leukemic cells from most patients with non-T-cell ALL and some patients with chronic myelocytic leukemia in blast crisis. J-5 antibody is not reactive with leukemic cells from patients with either T-cell ALL, chronic lymphocytic leukemia, acute myeloblastic leukemia, or stable-phase chronic myelocytic leukemia. In addition to diagnostic applications, anti-common ALL antigen monoclonal antibody has been used to study the effect of specific serotherapy in the treatment of ALL. In the first patient with ALL treated with J-5 antibody, a 90% reduction in circulating lymphoblasts occurred within 1 hr of starting antibody infusion. Despite continued serotherapy, however, lymphoblasts persisted in peripheral blood and bone marrow. Analysis of cell surface markers during serotherapy suggested that resistance to antibody-mediated lysis in vivo may have been due to antigenic modulation of leukemic cells in response to J-5 monoclonal antibody.

Elimination of malignant clonogenic cells from human bone marrow using multiple monoclonal antibodies and complement.

A clonogenic assay has been developed that utilizes Burkitt's lymphoma tumor cell lines to detect elimination of up to 5 logs of tumor cell contamination within human bone marrow. Different Burkitt's lymphoma lines bear one or more of a group of markers, including common acute lymphoblastic leukemia antigen gp26 (glycoprotein with a molecular weight of 26,000), B1, surface membrane immunoglobulin, HLA, beta 2-microglobulin, and Ia. Burkitt's tumor cells of the Namalwa line have been mixed with a 20-fold excess of irradiated human bone marrow cells. After treatment with one or more monoclonal antibodies and rabbit complement (RC), mixtures have been grown on a monolayer of irradiated human bone marrow cells and tumor cells enumerated by limiting dilution. Multiple treatments with antibody and RC were more effective than a single treatment in destroying clonogenic tumor cells which bore relevant determinants. Human serum components inhibited the lytic activity of RC in the presence of murine monoclonal antibodies. The total concentration of bone marrow cells proved critical in determining the complete elimination of tumor. Incubation of the Namalwa tumor cell line with RC and the J2 anti-gp26 eliminated more than 3 logs of malignant cells from a 20-fold excess of human bone marrow. Combinations of two monoclonal antibodies were more effective than any single antibody in eliminating Namalwa cells. A combination of three monoclonal reagents was no more effective than a combination of J2 and B1 or J2 and J5 in eliminating Namalwa cells. Treatment of human bone marrow with three antibodies and RC did not, however, produce a selective loss of nonmalignant GM-CFU-C, CFU-E, or BFU-E.

In vitro and in vivo killing of acute lymphoblastic leukemia cells by L-asparaginase.

L-Asparaginase (ASNase) is a potent antileukemic enzyme routinely used in the treatment of children with acute lymphoblastic leukemia. As part of investigations of the biological activity of ASNase, we have developed techniques which measure the in vitro and in vivo cell killing ability of ASNase. To study the effect of ASNase on in vitro survival of primary lymphoblasts, bone marrow mononuclear cells obtained from untreated patients with acute lymphoblastic leukemia were cultured with and without ASNase. After 5 days, viable cells were counted using trypan blue exclusion to calculate total cell kill due to ASNase. Propidium iodide exclusion, leukemia cell surface antigens, and flow cytometry were used to determine leukemia cell kill due to ASNase. Comparison of leukemia cell kill and total cell kill showed a direct linear relationship (n = 24, r = 0.7), preferential killing of leukemia cells by ASNase (slope = 0.66), and that use of leukemia cell surface markers yielded a more accurate measurement of leukemia cell killing. ASNase at concentrations from 0.0001 to 0.1 IU/ml had equal effects on extent of leukemia cell killing (P = 0.3 to 0.7), suggesting the absence of a dose response at the ASNase concentrations tested. As a measure of the in vivo response to ASNase treatment, the number of viable bone marrow leukemia cells in the patient prior to and 5 days after treatment with ASNase was measured as the product of (% of rhodamine 123 fluorescent [viable] cells) x (absolute leukemic infiltrate). The change which occurred in the viable leukemic infiltrate was the same for patients whether they received 25,000 or 2,500 IU/m2 of ASNase as a single drug. There was a linear correlation (n = 8, r = 0.9) between in vivo and in vitro leukemia cell killing by ASNase. Thus, the in vitro assay described here can be used to predict in vivo sensitivity to ASNase in acute lymphoblastic leukemia.

Elimination of leukemic cells from human bone marrow using monoclonal antibody and complement.

Human leukemic cells which bear the common acute lymphoblastic leukemia antigen can be lysed with a murine monoclonal antibody (J-5) in the presence of rabbit complement. Conditions have been defined for eliminating 51Cr-labeled common acute lymphoblastic leukemia antigen-positive NALM-1 cells or cryopreserved leukemic lymphoblasts from a 100-fold excess of human bone marrow. Optimal lysis is obtained with treatment for a total of 90 min. Three treatments for 30 min are more effective than two treatments for 45 min or one treatment for 90 min. Separation of marrow on a Ficoll:diatrizoate gradient does not permit more effective elimination of leukemic cells. Tumor cell lysis is inhibited by high concentrations of common acute lymphoblastic leukemia antigen-positive cells (2 X 10(7)/ml) and by high concentrations of bone marrow (10(8)/ml). Under optimal conditions, greater than 99% of 51Cr-labeled leukemic lymphoblasts can be eliminated from a 100-fold excess of human marrow. Selective removal of leukemic cells from human bone marrow in vitro should facilitate trials of autologous marrow transplantation.

Levels of 2'-deoxycoformycin, adenosine, and deoxyadenosine in patients with acute lymphoblastic leukemia.

2'-Deoxycoformycin (dCF), a potent inhibitor of adenosine deaminase, has recently undergone Phase I clinical trials and has been found to be therapeutically active in acute lymphoblastic leukemia. In this report, levels of dCF in plasma, plasma concentrations of adenosine and deoxyadenosine, and urine levels of deoxyadenosine were measured in leukemic patients undergoing treatment with dCF during a Phase I clinical trial. dCF was administered i.v. at a dose of 0.25 to 1.0 mg/kg (7.5 to 30 mg/sq m) for 3 consecutive days. Plasma drug levels of 2 to 6 microM were observed following the third dose of dCF, and drug accumulation occurred only at the 1-mg/kg dosage. In this limited series of patients, the plasma concentrations of adenosine and deoxyadenosine and the urine concentrations of deoxyadenosine did not show an obvious correlation with dCF dose, therapeutic response, or toxicity.

Phase I study of 2'-deoxycoformycin in acute lymphoblastic leukemia.

2'-Deoxycoformycin (2'-dCF), a tight-binding inhibitor of adenosine deaminase, was administered to 26 pediatric patients with acute lymphoblastic leukemia in a Phase I study. Doses ranged from 0.25 to 1.0 mg/kg given i.v. for 3 consecutive days. Common toxicity included nausea, vomiting, diarrhea, hepatocellular enzyme elevations, and conjunctivitis. Lymphopenia occurred in all patients. The most serious adverse effects were acute tubular necrosis and central nervous system toxicity, which appeared to be dose related. In addition, two patients given the 0.75-mg/kg dose developed severe hepatic toxicity, although this could not be ascribed definitively to 2'-dCF. Antitumor activity was observed in eight patients, two of whom experienced a complete remission. Inhibition of lymphoblast adenosine deaminase activity was noted in the majority of cases and was observed at all doses. Antileukemic activity occurred at doses of 2'-dCF which were not associated with limiting toxicities. These results suggest that 2'-dCF is active against acute lymphoblastic leukemia and that a starting dose of 0.5 mg/kg/day be utilized in Phase II studies.

Clinical pharmacology of 9-beta-D-arabinofuranosyladenine in combination with 2'-deoxycoformycin.

It has been suggested that, by inhibiting the adenosine deaminase (ADA)-mediated catabolism of 9-beta-D-arabinofuranosyladenine (ara-A), 2'-deoxycoformycin (DCF) would increase the half-life (t1/2) of ara-A, a compound with known antileukemic activity. To test this hypothesis, we collected serial plasma samples from five patients with refractory acute lymphoblastic leukemia who participated in a Phase I trial of i.v. DCF 915 mg/sq m) in combination with i.v. single-dose ara-A (120-250 mg/sq m). In four of these patients, of whom three were known to have achieved greater than 98% ADA inhibition, a mean ara-A t1/2 of 227 min was achieved. Extrapolated peak levels (i.e., following infusion of ara-A) ranged from 1.5 to 7.4 micrograms/ml (mean, 4.2 micrograms/ml). Elimination of drug appeared to follow a single-compartment model. In two patients who received ara-A without prior DCF and in a third patient who had significant residual ADA activity despite DCF, ara-A was unmeasurable within 5 min of the end of infusion. These data confirm that the kinetics of ara-A catabolism can be altered by inhibition of ADA and suggest that more than one dose of DCF may be necessary for complete inhibition of the enzyme and optimal pharmacological modulation of ara-A.

Pediatric phase I trial and pharmacokinetic study of tiazofurin (NSC 286193).

Tiazofurin (2-beta-D-ribofuranosylthiazole-4-carboxamide), a new nucleoside antimetabolite, was evaluated in a phase I trial involving children with refractory cancers. The drug was administered i.v. as a 10-min infusion daily for 5 consecutive days repeated at 3-week intervals. The dose ranged from 550 to 3300 mg/sq m/day. Seventeen patients received 23 courses and were evaluable for toxicity. The maximally tolerated dose was 2200 mg/sq m/day. The major dose-limiting toxicities were nonhematological. Neurotoxicity, including headache, drowsiness, and irritability, was common and was the principal dose-limiting toxicity at the higher doses. Severe myalgias were also dose limiting in one patient. Other side effects were mild, reversible elevations in serum transaminases; nausea, vomiting, and diarrhea; mild hypertension; dysphagia; and exfoliative dermatitis of the hands and feet. Myelotoxicity was not significant. The pharmacokinetics of tiazofurin was studied in 16 patients. Plasma disappearance was triphasic with half-lives of 9.7 min, 1.6 h, and 5.5 h. Clearance was dose related, ranging from 120 ml/min/sq m at 550 mg/sq m/day to 70 ml/min/sq m at 3300 mg/sq m/day. The primary route of elimination was renal with 85% of the drug recoverable in the urine as the parent compound in the 24 h following administration.