* Institute of Microbiology, Milan, Italy;
CIEMAT, Madrid, Spain;
BIBRA International, Carshalton, United Kingdom;
ECVAM, Institute for Health and Consumer Protection, Joint Research Centre, European Commission, Ispra, Italy;
¶ Wayne State University, Detroit, Michigan 48202;
| ESMISAB, Plouzanè, France; and
|| VITO, Mol, Belgium
Received June 3, 2003; accepted June 25, 2003
ABSTRACT
In a previous study of prevalidation, a standard operating procedure (SOP) for two independent in vitro tests (human and mouse) had been developed, to evaluate the potential hematotoxicity of xenobiotics from their direct and the adverse effects on granulocyte-macrophages (CFU-GM). A predictive model to calculate the human maximum tolerated dose (MTD) was set up, by adjusting a mouse-derived MTD for the differential interspecies sensitivity. In this paper, we describe an international blind trial designed to apply this model to the clinical neutropenia, by testing 20 drugs, including 14 antineoplastics (Cytosar-U, 5-Fluorouracil, Myleran, Thioguanine, Fludarabine, Bleomycin, Methotrexate, Gemcitabine, Carmustine, Etoposide, Teniposide, Cytoxan, Taxol, Adriamycin); two antivirals (Retrovir, Zovirax,); three drugs for other therapeutic indications (Cyclosporin, Thorazine, Indocin); and one pesticide (Lindane). The results confirmed that the SOP developed generates reproducible IC90 values with both human and murine GM-CFU. For 10 drugs (Adriamycin, Bleomycin, Etoposide, Fludarabine, 5-Fluorouracil, Myleran, Taxol, Teniposide, Thioguanine, and Thorazine), IC90 values were found within the range of the actual drug doses tested (defined as the actual IC90). For the other 10 drugs (Carmustine, Cyclosporin, Cytosar-U, Cytoxan, Gemcitabine, Indocin, Lindane, Methotrexate, Retrovir, and Zovirax) extrapolation on the regression curve out of the range of the actual doses tested was required to derive IC90 values (extrapolated IC90). The model correctly predicted the human MTD for 10 drugs out of 10 that had "actual IC90 values" and 7 drugs out of 10 for those having only an extrapolated IC90. Two of the incorrect predictions (Gemcitabine and Zovirax) were within 6-fold of the correct MTD, instead of the 4-fold range required by the model, whereas the prediction with Cytosar-U was approximately 10-fold in error. A possible explanation for the failure in the prediction of these three drugs, which are pyrimidine analogs, is discussed. We concluded that our model correctly predicted the human MTD for 20 drugs out of 23, since the other three drugs (Topotecan, PZA, and Flavopiridol) were tested in the prevalidation study. The high percentage of predicitivity (87%), as well as the reproducibility of the SOP testing, confirm that the model can be considered scientifically validated in this study, suggesting promising applications to other areas of research in developing validated hematotoxicological in vitro methods.
Key Words: GM-CFU assay; acute neutropenia; maximum tolerated dose; phase I trial; myelotoxicity.
In vitro models of hematopoiesis are being used increasingly in investigative hematopathology and in preclinical safety studies on candidate drugs (Deldar, 1994; Deldar and Parchment, 1997
; Deldar and Stevens, 1993
). These models are also useful for determining the relative sensitivities of various animal species to haematotoxic effects and for studying synergistic and antagonistic effects of several compounds (Du et al., 1990
). The type of hematotoxicity most frequently and most thoroughly studied in vitro is the acute effect of toxicants on bone marrow progenitors, such as granulocyte-macrophages (CFU-GM), erythroids (CFU-E), and megacaryocytes (CFU-MK), which is quantified from the number of surviving progenitors as a function of exposure level under maximally stimulatory cytokine concentrations (Metcalf, 1984
). Since haematotoxicity can result from either the direct interference of the toxicant with the different haematopoietic progenitors or with the expression of cytokines and their receptors, many different protocols have been developed and proposed for in vitro hematotoxicity testing (Gribaldo et al., 1996
, 1998
; Lewis et al., 1996
; Naughton et al., 1992
; Noble and Sina, 1993
; Parchment et al., 1998
; Parent-Massin and Thouvenot, 1995
; Parent-Massin et al., 1993
; Pessina, 1998
; Pessina et al., 1999
; San Roman et al., 1994
; Schoeters et al., 1995
; Van Den Heuvel et al., 1997
). All these tests are modifications of the original technique suggested by Bradley and Metcalf (1966)
and then developed and modified by other authors (Dexter and Spooncer, 1987
; Dexter and Testa, 1976
; Dexter et al., 1973
). In vivo substances such as antineoplastics, microbial toxins, and ionizing radiation (Bruce et al., 1966
; Grande and Bueren, 1995
) destroy the rapidly dividing marrow progenitors, and a single exposure can result in acute, reversible neutropenia or thrombocytopenia 4 to 20 days later. A rapid repopulation of the progenitor compartment precedes the recovery of peripheral counts by several days (Greenberg et al., 1974
; Neelis et al., 1997
). Important goals during preclinical drug development is to predict whether a new agent will be clinically toxic to the bone marrow, whether the toxicity will be specific to one cell lineage (lymphocytes, neutrophils, megakaryocytes, or erythrocytes), at what dose or plasma level the drug will be toxic, which model best predicts the clinical situation, and when the onset and nadir of cytopenia and the onset of recovery will be likely to occur. Myelotoxicity is one of the major limitations to the use of full doses of antitumor agents, and the goal in the regulatory setting usually emphasizes the prediction of two levels of exposure: the highest dose that will not cause a clinically adverse effect and the dose that causes maximally tolerated, reversible perturbations in peripheral blood counts, termed the maximum tolerated dose (MTD). Selection of the starting dose for the phase I trial, which is typically based on the MTD of the most sensitive species, is critical, in that the dose selected must not be toxic, while at the same time it must be high enough to give the patients therapeutic benefit. Even with the advent of the use of molecular targeting to develop new therapeutic approaches, the majority (>50% in the last 15 years) of anticancer drugs still produce myelosuppression as the dose-limiting toxicity (DLT) in humans (Parchment, 2000
; Parchment et al., 1998
).
In vitro tests could refine safety margins by reducing toxicological uncertainties due to animal/human extrapolation, and would provide a more rational basis for calculating clinical dosages and for setting human exposure limits. With anticancer drugs, in vitro studies should be undertaken to identify those compounds that are significantly more toxic to humans than to either dogs or rodents. By identifying such compounds, it would be possible to decrease the risk of a lethal overdose in the first cohort of patients to which they are administrated, a risk that cannot be identified during current preclinical testing strategies. An in vitro assay could highlight the potency difference between humans and the preclinical test species, so that the starting dose in phase I clinical trials could be considerably closer to the MTD, without compromising safety. Thus, not only would phase I clinical trials be completed more quickly, but fewer patients would be treated with ineffective doses. In this respect, the predictivity of the data obtained from animal studies could be increased by in vitro tests, and the level of uncertainty concerning human safety could be decreased (Grande and Bueren, 1995). Validated in vitro tests for hematotoxicity could help to answer some of the above-mentioned questions, and contribute to a reduction in the number of animals required in preclinical toxicology (Balls et al., 1995
; Curren et al., 1995
).
As reported in Parchment et al. (1994), a correlation was found between the severity of neutropenia in the clinic with pyrazoloacridine and the inhibition of CFU-GM in vitro. Subsequently, in vitro/in vivo correlations were found for the camptothecins (Erickson-Miller et al., 1997
) and anguidine (Parent Massin and Parchment, 1998
). A key finding in these studies was that the concentration that inhibited CFU-GM by 90% (IC90) was a more predictive endpoint for the MTD in animals and in humans than the IC50 (Parchment et al., 1997
, 1998
). The success to date lies primarily in the identification of the in vitro inhibition concentration that can predict the MTD. In a previous prevalidation study (on six anticancer drugs) a standard operating procedure (SOP) for murine and human CFU-GM was developed. Reproducible IC90 values obtained with SOP were used to predict human MTDs, which were compared to the actual human MTD (Pessina et al., 2001
). In this paper, we describe an international validation study supported by the European Centre for the Validation of Alternative Methods (ECVAM) designed to evaluate the predictive capacity of this model when applied to clinical neutropenia, by testing an additional 20 drugs. The prediction model adopted utilizes information from the in vitro analysis of toxic effects on the actual human target cell, and offers the advantage of being mechanism-naïve and would only fail to identify hematotoxicants that adversely affect myelopoiesis via indirect physiological mechanisms such as induced release of inhibitory cytokines, inhibited release of stimulatory cytokines, or metabolic activation of pro-toxicants. It provides human toxicology and pharmacologic information in the laboratory setting and an experimental basis for selecting the best animal models for investigating clinical haematotoxicity.
MATERIALS AND METHODS
Methylcellulose Culture Medium
The medium for murine and human cells was prepared by StemCell Technologies (Vancouver, BC, Canada) according to specific modifications of the Methocult, as suggested by the management team of the validation study (Pessina et al., 2001). The medium contained 1% methylcellulose in IMDM (Iscoves Modified Dulbeccos Medium), 30% fetal bovine serum (FBS), 1% bovine serum albumin, 2- mM L-glutamine, and 10-ng/ml granulocyte/macrophage-colony stimulating factor (human-rec-GM-CSF or murine-rec-GM-CSF). A unique batch of each medium was prepared, and aliquots of the same batch were supplied frozen to the laboratories.
Haematopoietic Progenitors
Human umbilical cord blood cells (hu-CBC) from five different donors were supplied frozen by Poietic Technologies (Gaithersburg, MD), according to a protocol approved by the Institutional Review Board (IRB), and all of the cryotubes were stored in liquid nitrogen. A sufficient number of aliquots of each CBC donor was distributed to each laboratory, so that all participants could work on the same batch of a cell preparation during each experimental phase.
Murine bone marrow cells were collected from male BDF/1 (C57Bl/6xDBA-2) mice, 812 weeks old, according to the procedure described in Pessina et al. (2001).
Drug Selection
Twenty drugs were selected according to their recognized or potential hematotoxicity in clinical use. As negative controls, drugs with no previously described hematotoxicity were selected. The main characteristics of the drugs selected are reported in Table 1. The list comprised 14 antineoplastics (Cytosar-U, 5-Fluorouracil, Myleran, Thioguanine, Fludarabine PO4, Bleomycin, Methotrexate, Gemcitabine, Carmustine, Etoposide, Teniposide, Cytoxan, Taxol, Adriamycin); two antivirals (Retrovir and Zovirax,); three other therapeutic indications (Cyclosporin, Thorazine, and Indocin); and one pesticide (Lindane).
|
Experimental Design
To evaluate the intralaboratory and interlaboratory variabilities of the in vitro assay, three laboratories (Brest, Milan, Mol) tested the drugs on the murine model (40,000 mu-BMC/dish), and three laboratories (Detroit, Ispra, Madrid) used the human model (75,000 hu CBC/dish).
The experimental design was optimized for testing three drugs (in triplicate) simultaneously on one day for each experiment. Each drug was tested three times: The first experiment was a screening test (ST) to determine the range of the drugs activity; the other two tests were performed to determine the inhibitory concentration (IC) by using a narrower range of drug concentrations as described below. For each experiment an internal control was set up to confirm the linear relationship between the cells seeded and the colonies scored. This control was based only on the assessment of two points (CTRL1 and CTRL3) of the curve studied in the prevalidation phase and corresponded to 2,500 and 40,000 cells/dish for the murine model and to 10,000 and 75,000 cells/dish for the human model (Pessina et al., 2001).
CFU-GM Assay
The detailed SOP was given in the report of the prevalidation study (Pessina et al., 2001). Here we describe only the modifications concerning drug dilution and the preparation of tubes, which were changed as a consequence of the "blind conditions" needed to perform the validation study. Briefly, 11 tubes of cell culture mixture were prepared for each experiment, according to the following experimental design: CTRL1 and CTRL3 (linearity controls), D0 (vehicle controls), and D1D8 for the doseresponse curve. To each tube containing 4.0 ml of methylcellulose culture medium were added 100 µl of IMDM (to CTRL1 and CTRL2) and 78 µl of IMDM (to D0D8). Then, 22 µl of the vehicle were added to D0 and 22 µl of each toxicant dilution to D1D8. Each tube was used to prepare three culture dishes. All of the toxicant dilutions were prepared at x200 the final dilution, in order to obtain the final fold dilutions of drug in the culture dish of 5 x 10-3, 10-3, 10-4, 10-5, 10-6, 10-7, 10-8, 10-9 from D8 to D1, respectively (the concentration of the vehicle was of 0.5% vol for each dilution).
The cultures were incubated at 37°C in air + 5% CO2 under saturated humidity for 7 days (murine assays) or 14 days (human assays). All dishes were scored for colony counts in a random fashion. The different types of colonies (compact colonies, diffuse and spread colonies, multicentric colonies, burst-forming units) were all counted as one colony. Since the humidity level is a critical parameter during incubation, the evaporation rate (ER) was determined as described in Pessina et al. (2001).
Passing from the Screening Test to IC Determination Phase
The GFU-GM results obtained in the Screening Test enabled the following to be identified: the first drug dilution that completely inhibited CFU-GM (FCID) and the last drug dilution that did not inhibit CFU-GM (LNID). Then the log dose differential between the LNID and FCID was calculated, as follows:
![]() |
To determine the new drug concentrations to use in the two assays of the IC determination phase, the FCID was assigned to D8 and the LNID to D2, the value was divided into six parts of size (
=
: 6), and these parts were assigned to the dilution levels Dn (D2D7) according the following formula:
![]() |
In order to reduce the variability, all of the dilutions were performed by starting from the D8 concentration of toxicant stock. However, under some conditions, the volume required for making the dilution was smaller than could be pipetted, so, for these concentrations, dilutions of the drug stock were prepared before making the working stock.
Statistical Analysis
At the end of the study, the data were collected by each laboratory in an Excel template, the test drugs were decoded, and the actual concentrations used in the assay were calculated. The statistical analysis of the data was performed by means of the Generalized Linear Model and PROBIT procedures of the SAS package (SAS Institute Inc., Cary, NC). For all of the comparisons, a 0.05 -type error was considered as significant. Three main steps were followed:
Verifying the linear relationship: Cells seeded-CFU-GM colonies.
The slope of the line between two internal controls (CTRL1 and CTRL3) was estimated by considering the ratios between CTRL3 and CTRL1 and then testing the null hypothesis that the slope was equal to zero. The homogeneity of results for each laboratory (and experiment) was verified by performing an analysis of variance (ANOVA). The percentage of coefficient of variability (CV %) at CTRL3 (calculated as standard deviation/mean x 100) was also considered.
Analyzing the best fitting model.
Several regression models were fitted to study the inhibition of CFU-GM in IC determination experiments. A goodness of fit test, based on the likelihood ratio, chi-square, was verified for each drug and for each species tested (human or murine), according to the drug concentration (expressed as µg/ml), and three cumulative distribution functions were used to model the response probabilities (NORMAL: normal distribution for the probit model; LOGISTIC: logistic distribution for the logit model; GOMPERTZ: Gompertz distribution for the gompit model). For each species (murine and human), drug, and laboratory, the best fitting model was selected and the sources of variability were identified for the whole curve.
Estimating the inhibitory concentrations.
Based on the best fitting model selected for each drug, the ICs were estimated from IC1 to IC99 by the probit procedure. Homogeneity of the inhibitory concentrations was analysed by studying both the interlaboratory and the intralaboratory variability by ANOVA.
Prediction Model
The prediction model for calculating the human MTD was based on the following algorithm:
![]() |
This prediction model, studied and applied in the prevalidation study (Pessina et al., 2001), correlates the inhibition of CFU-GM in vitro and the depth of the absolute neutrophil count (ANC) nadir in vivo. Because pharmacokinetic differences across species may contribute to as much as a 4-fold difference in MTD, and the prediction model does not accommodate this source of variability, we considered an "accurate prediction" as the prediction of a human MTD that lies within 4-fold of the actual human MTD value (Erickson-Miller, 1997
; Parchment et al., 1993
, 1994
; Volpe et al., 1996
).
RESULTS
Internal Linearity Control
The absolute counts of CFU-GM in CTRL3 controls were 81.7 ± 53.4 for human and 83.8 ± 46.6 for murine colonies. The ratios CTRL3 /CTRL1 gave means of 13.5 ± 0.9 for the murine and 9.2 ± 0.4 for the human model. The linearity and proportionality of response are independent of the absolute number of colonies that each laboratory obtained in D0 dishes (which can differ greatly). Only one test out of 300 had to be repeated because the linearity test failed and produced unacceptable results (human model with CTRL3/CTRL1 ratios < 7).
The statistical analysis of the internal linearity controls verified the required linearity and proportionality between the number of cells seeded and colonies counted, confirming a good correlation with high coefficient of determination values of R2 = 0.45 in the human model and R2 = 0.57 in the murine model. No significant intralaboratory and interlaboratory variability was observed (human, p = 0.60; murine, p = 0.18), and the regression showed a very significant slope (p = 0.0001). On the basis of these results, all of the experiments performed were considered to meet the acceptance criteria, thereby confirming the good performance of the protocol observed previously in the prevalidation study.
IC Determination
For over 90% of the drugs, both in human and murine, the best fitting model was compatible with the normal or logistic distribution, and since the chi-square values did not differ significantly between them, the following analysis was performed by considering, for all of the drugs, the normal distribution. To normalize the doseresponse curve, all of the regression lines were determined by considering as 100% the absolute counts of CFU-GM scored in D0 (vehicle control).
As is shown in Tables 2 and 3
, for some drugs, all of the IC values were determinable within the range of the actual concentrations tested, which were calculated by interpolation on the fitted regression curves. For the other drugs, the IC values had to be extrapolated out of the range of drug concentrations tested. For these drugs the reliability of the extrapolated IC values strongly depends on the maximal experimental IC actually determined by the assay. Therefore, the degree of reliability of the IC90 was predicted on the basis of the maximal common IC (McIC) that the test was able to estimate both in the murine and human models within the range of the actual drug concentrations in three laboratories (McIC > 50: acceptable reliability; McIC < 50: poor reliability).
|
|
The IC90 value was interpolated, within the range of doses tested, for 10 drugs (Adriamycin, Bleomycin, Etoposide, Fludarabine, 5-Fluorouracil, Myleran, Taxol, Teniposide, Thioguanine, and Thorazine) by three laboratories, for two drugs (Cytoxan and Gemcitabine) by two laboratories, for two drugs (Cytosar-U and Retrovir) by one laboratory. For the remaining six drugs (Indocin, Lindane, Methotrexate, Zovirax, Carmustine, and Cyclosporin), the IC90 could not be estimated. Based on the McIC value determinated in each laboratory, the reliability of the IC90 values extrapolated from the regression curve was acceptable for five drugs (Carmustine, Cytosar-U, Cytoxan, Gemcitabine, and Retrovir), since their McIC > IC50. For the other drugs the extrapolation of IC90 values is likely to be imprecise, because the McIC was lower than the IC50.
The ANOVA showed a high intralaboratory homogeneity (only Adriamycin for IC50 values showed a p < 0.05). A greater heterogeneity for four drugs was observed for interlaboratory variability: IC50 (Adriamycin and Methotrexate) and IC50 and IC90 (Myleran and Taxol). However, this significant interlaboratory variation affected the IC90 values, and therefore the prediction model, for only two drugs.
These results clearly indicate that, in the determination of IC, the laboratory represented a significant source of variability for 4 drugs out of 20 (corresponding to 20% of the drugs tested) and for 6 IC determinations out of 40 (15%).
Murine model.
The analysis of data from the murine model showed a high degree of homogeneity (Table 3). For 15 drugs (Adriamycin, Bleomycin, Cyclosporin, Cytosar-U, Etoposide, Fludarabine, 5-Fluorouracil, Gemcitabine, Methotrexate Myleran, Taxol, Teniposide, Thioguanine, Thorazine, and Zovirax), the IC50 values were determinable by three laboratories and for one drug (Carmustine) by two laboratories. For three drugs (Indocin, Lindane, and Retrovir), IC50 was not estimable by any laboratory.
The IC90 values were determinable by three laboratories for 12 drugs (Adriamycin, Bleomycin, Cytosar-U, Etoposide, Fludarabine, 5-Fluorouracil, Gemcitabine, Myleran, Taxol, Teniposide, Thioguanine, and Thorazine), by two laboratories for Cytoxan, and by one laboratory for Retrovir. For six drugs (Indocin, Lindane, Methotrexate, Zovirax, Carmustine, and Cyclosporin), none of the laboratories was able to determine the IC90 value.
The analysis of the IC values determinable by each laboratory suggested that the extrapolated IC90 values for Carmustine, Cyclosporin, Cytosar-U, Gemcitabine, and Zovirax had a sufficient degree of reliability with the McIC > 50%. For the remaining drugs, the extrapolated IC90 values were imprecise because the McIC was lower than the 50%.
A very high intralaboratory homogeneity was observed (only the IC90 determination for Teniposide showed a p < 0.05). The interlaboratory variability was lower than that observed in the human model and concerned five drugs: IC50 (Fludarabine), IC90 (Etoposide), IC50 and IC90 (Cyclosporin, Myleran, and Teniposide). As in the human model, the interlaboratory variability in the murine study influenced the IC determination for 5 drugs out of 20 (25%) and 8 IC determinations out of 40 (20%).
In the human model, the interlaboratory variation observed seems have been due to the use of different donors rather than to CBC variations in methodology. In fact, the very rare occurrence of intralaboratory variation during the human CFU-GM study, and the normalization of the colony counts to the D0 control, suggest that biological variation in the CBC response across individual donors may contribute more. In the murine model, the interlaboratory variation observed seems more attributable to the technical methodology, because the bone marrow was obtained from syngenic mice.
Prediction of Human MTD
The IC90 values for the murine and human CFU-GM generated with our SOP were applied to the algorithm of the prediction model according to two different groups of drugs, as reported in Tables 4 and 5
.
|
|
Table 5 shows the IC90 ratios for the 10 drugs that required extrapolation of IC90 values from the regression curve outside the dose range tested. The prediction model applied to this group of drugs "accurately" predicted 7 drugs out of 10 (70%) but failed to accurately predict the human MTDs of Cytosar-U, Gemcitabine, and Zovirax.
DISCUSSION
General Considerations on the MTD for Neutropenia
The IC90 value is the concentration that inhibits CFU-GM colony formation by 90%, and half of the tested compounds were potent enough to inhibit both the human and murine CFU-GM to this degree within the range of concentrations that were completely soluble in water.
Severe toxicity in this context means life-threatening toxicity, in which supportive medical care may be required for recovery and in which the marrow MTD will equal the MTD for the human being as a whole. In cases where hematopoiesis is less sensitive to the toxicant than another organ system, the human MTD will be lower than the bone marrow MTD. However, it is important to note that the human MTD will never be higher than the bone marrow toxicity, at least as irreversible in the remaining life span of the individual. Whether a toxicant will cause neutropenia depends entirely on whether the other organ systems can tolerate this exposure level. Therefore, according to these above-mentioned concepts, our validated prediction model predicts the human dose that will be associated with life-threatenig neutropenia.
It is also important to note the fundamental departure of this validation study from the customary toxicology strategy of hazard classification and the importance of this departure to the success of the study. Hazard classification depends on finding some in vitro characteristic that is highly associated with the toxicity of interest and not with other outcomes. At the time this study was planned, it was widely appreciated that short-term exposure to many toxicants does not necessarily cause hematoxicity in vivo, despite the fact that inhibiting proliferation of the neutrophil precursor in the bone marrow leads to reversible neutropenia. This reasoning led to the somewhat obvious conclusion that the level of exposure relative to a toxicants potency is the determining factor for the clinical outcome, rather than the potency or degree of in vitro inhibition per se.
Therefore, all of the compounds must be classified as potentially hematotoxic, and the only issue to be predicted by in vitro alternatives is the dose level that will affect the bone marrow.
Abandoning hazard classification led to the realization that the concept of negative and positive controls was not useful for the validation process. Instead, the study required a set of test compounds that reflected a range of potencies at the marrow targets and the widest possible range of interspecies differences in tolerated doses between mouse and human. "Negative control" in this context means a compound that does not cause hematotoxicity in vivo because its substantially greater potency on another organ system prevents exposure levels from becoming high enough to affect the bone marrow. This argument provided the rationale for the dependence of the validation study on the use of oncology drug products as test articles. This set of compounds has two additional advantages as test substances: There are complete data sets on their murine and human toxicology from product safety studies and controlled clinical trials, respectively, and it is ethical to obtain human dosetoxicity relationships, including life-threatening exposure levels, in the context of closely monitored clinical trials. Almost all of the oncology products tested in the current study can be administered to humans at exposure levels that are myelosuppressive, because they have narrow therapeutic indices that require toxicity to achieve efficacy, and for most of them neutropenia is the most common dose-limiting effect. The exceptions are Bleomycin, which causes dose-limiting pulmonary toxicity without myelosuppression, and probably Cytoxan, which is converted to the active 4-hydroxycyclophosphamide metabolite and other metabolites that may contribute to its toxicity. The toxicity profile of Cytoxan itself in the absence of metabolic activation is not known. Clinical trials have shown that antivirals such as Retrovir (zidovudine, "AZT") and Zovirax (aciclovir) also cause dose-limiting toxicity to hematopoiesis. However, they are usually used at dosages that do not cause neutropenia, because their therapeutic indices are higher than those of oncology products. The remaining test compounds reach myelosuppressive levels only in the context of overdosage (Thorazine), or they were the validation studys negative controls (Cyclosporin and Indomethacin), the exposure levels of which could never reach myelosuppressive levels, because of severe organ toxicities occurring outside the bone marrow. This group of test items met the need for a very wide range of potencies at the marrow target in humans, including compounds that have such a low potency that the marrow MTD will be substantially greater than the human MTD. They also represent a wide range of mechanisms of action, structural diversity, and biophysical properties.
Considerations on the Validated CFU-GM SOP
This prevalidation-validation effort incorporated several key technical aspects not usually found in the common method of CFU-GM testing that were probably critical to the success of this study.
First, several chemical modifications to the Stem Cell MethoCult culture medium were made to avoid substances that could interfere with the toxicity of the test articles. 2-Mercaptoethanol is usually found in this culture medium because it improves colony number and colony size, but, being a mercaptan and nucleophilic, it could react with many of the chemically reactive toxicants that might need to be evaluated in the CFU-GM test. Likewise, transferrin is a key component in culture media that support cell proliferation, but the ferrous iron that it carries, in the presence of the high oxygen tension that exists in most CO2 incubators, can react with xenobiotics to produce free oxygen radicals. Among the test substances in this study, the iron-catalyzed generation of free radicals has been implicated in the toxicity of Bleomycin and Adriamycin. The potential of 2-ME and the ferrous iron to react with toxicants and thereby lower their effective exposure levels in the culture medium was only of minor concern, because the human and murine culture systems would have contained the same levels of 2-ME and transferrin; therefore, the relative measure of potency across species required by the prediction model would not have been affected. However, the management team was concerned about the products of these chemical reactions modifiying a test articles toxicity or exhibiting biological activity themselves in the CFU-GM test system and then misinterpreted this artifactual activity as that of the test compound.
Second, the growth stimulant of colony formation was limited to recombinant GM-CSF. Cell-line conditioned medium and cytokine/growth factor cocktails were avoided, because they stimulate colony formation by additional neutrophil precursor cells, which are less mature than the CFU-GM. The role of these immature cells in hematotoxicity is not yet clear, but in vivo modeling in mice has shown that they are probably not targets for acute toxicant exposure that causes reversible neutropenia. It is not possible to know whether the assay could have been validated without making these changes.
Third, the mouse MTD values used for prediction modeling, and the human MTD values to which they were compared, were derived from studies with the exact same route of toxicant exposure (usually intravenous, but also including oral) and similar dosing regimens. Given the schedule dependency of many marrow toxicants, it was very important to match route and regimen across species, so the true interspecies difference measured by the CFU-GM test, which is a difference in potency at the same target organ, could be detected accurately. Comparing, for example, one-day dosing in the mouse versus five consecutive days in the human can hide the true interspecies variation in susceptibility underneath an overwhelming effect of scheduling on toxicity.
Consideration of Model Predictivity
This validation study, performed with a panel of 20 drugs, confirms that the SOP developed for assaying human and murine GM-CFU generates reproducible IC90 values that were applied to our model for predicting acute systemic doses that will cause severe, reversible neutropenia in treated patients (marrow MTD). Because the intralaboratory variability was lower than the interlaboratory one, the best homogeneity for IC determination is attained when human and murine CFU-GM assays are performed in the same laboratory. The biological differences due to different donor cord blood cells is likely to contribute to the greater interlaboratory variation in human than mouse CFU-GM data. However, this variation was acceptable and did not compromise the predictive accuracy of the model. For 10 drugs, IC90 values were found within the range of the actual drug doses tested (defined as the actual IC90). However, for the other 10 drugs, extrapolation on the regression curve out of the range of actual doses tested was required to derive IC90 values (defined as the extrapolated IC90). Our method correctly predicted the MTD for 10 drugs that had actual IC90 values.
Lindane and Indocin were selected as negative controls, and they indeed were nontoxic to CFU-GM, failing to reach even predicted IC 90 values because of a general lack of CFU-GM in vitro. However, relatively accurate IC50 values were available from the human CFU-GM assay; therefore we used a secondary prediction model that stated that the peak (maximum) plasma concentration of any toxicant that did not cause acute neutropenia in humans will lie below its IC50 value in the CFU-GM assay. Note that this model does not involve interspecies comparisons, and the IC50 value was chosen because this seemed to be the greatest amount of CFU-GM loss that is not associated with neutropenia in a small number of studies (Parchment et al., 1994; Parent-Massin and Parchment, 1998
). For Indocin, 10.815.7 (avg 13.5) mcg/ml plasma concentration of drug causes severe toxicity in just 7% of patients, and it is usually not bone marrow suppression. These plasma concentrations lie below the IC50 value of 264 mcg/ml in the human CFU-GM assay, so this toxicants risk of neutropenia was correctly predicted by the model. Several h after acute Lindane exposure, serum concentrations >0.2 mcg/ml cause seizure, >0.5 mcg/ml cause myonecrosis, and >1.2 mcg/ml cause death (Aks et al., 1995
; Davies et al., 1983
; Starr and Clifford, 1972
). However, these cases of accidental exposure were not associated with any reported bone marrow suppression, even though an extensive emergency room workup was performed in each case to characterize the extent of overdose. The lethal plasma concentration of 1.2 mcg/ml lies far below the IC50 value of 188 mcg/ml in the human CFU-GM assay, so this toxicant is counted as a success in the performance of the prediction modeling.
It is very important to recognize that the lack of in vitro CFU-GM toxicity provides evidence that the assay specifically detects myelosuppressive agents. Xenobiotics that do not usually cause hematopoietic toxicity failed to inhibit the neutrophil progenitor.
For those having only an extrapolated IC90, the method correctly predicted the MTD for seven drugs. Two of the incorrect predictions were within 6-fold of the correct MTD, instead of the 4-fold range required by the model. The final incorrect prediction for Cytosar-U was about 10-fold in error. However, it is important to note that these incorrect predictions underestimated the human MTD, and therefore, from the perspective of product safety, these predictions would have overestimated human risk rather than underestimated it (a more serious mistake) (Collins et al., 1990).
It remains to be explained why the method failed to predict the MTDs for Cytosar-U, Gemcitabine, and Zovirax. It seems realistic to suspect that this was due to IC90 "extrapolation" out of the range of the actual doses tested. Given the simplicity of the prediction model, underprediction is likely to originatefrom two sources. One explanation could be the irregular-shaped doseresponse curves that deviate substantially from the linear extrapolations used in the analysis to estimate IC90 values. Rather, these relationships may curve downward to more potent IC90 values than predicted. Second, three drugs with erroneous predictions are pyrimidine analogues, and it is possible that differences in the levels of endogenous natural pyrimidines that antagonize drug toxicity have not been accurately modeled in vitro.
Three drugs that were tested in the prevalidation study were not included in the validation phase: Topotecan, PZA, and Flavopiridol. The method also correctly predicted the human MTD for these drugs during that study. Therefore, when considered in total, the method and the prediction model correctly predicted the human MTD for 20 drugs out of 23 (87%). Coupled with the reproducibility of the SOP application, this 87% predictivity and the 94% predictivity for nonnucleoside structures (15 drugs of 16) confirm that the SOP and the prediction model can be considered scientifically validated in this study.
This favorable outcome suggests promising areas for further research in developing validated hematotoxicology tests: (1) verification of the introduction of naturally occurring pyrimidines and purines into the culture medium at physiological levels; (2) application of the SOP in a microtest (96-well plates) for high throughput screening of compounds; and (3) extension of this SOP to the rat and the dog. The greatest reduction and refinement in rodent and nonrodent use in toxicology will occur in these species, which are used throughout the world for product safety testing. Canine CFU-GM does not grow well in response to either murine or human GM-CSF, so a source of this cytokine will be required before the validation of CFU-GM assays for the dog will be possible.
ACKNOWLEDGMENTS
Financial support for this prevalidation study was provided by the European Commission (ECVAM, IHCP, Joint Research Centre), under the terms of Contract No.12230-96-10F1ED ISPI. The authors thank Dr. Andrew Worth (ECVAM) for critically reviewing the manuscript.
NOTES
1 To whom correspondence should be addressed at Institute of Microbiology, Via Pascal 36, 20133 Milan, Italy. Fax: +39 2 50315068. E-mail: augusto.pessina{at}unimi.it.
REFERENCES
Aks, S. E., Krantz, A., Hryhrczuk, D.O., Wagner, S., and Mock, J. (1995). Acute accidental lindane ingestion in toddlers. Ann. Emerg. Med. 26, 64751.[ISI][Medline]
Anderson, H., Thatcher, N., Walling, J., and Hansen, H. (1996). A phase I study of a 24 h infusion of gemcitabine in previously untreated patients with inoperable non-small-cell lung cancer. Br. J. Cancer 74, 460462.[ISI][Medline]
Anonymous (1970). Antirheumatic drugs: Effects and toxicity. Arthritis Rheum. 13, 598603.[Medline]
Balls, M., Blaauboer, B. J., Fentem, J. H., Bruner, L., Combes, R. D., Ekwall, B., Fielder, R. J., Guillouzo, A., Lewis, R. W., Lovell, D. P., et al. (1995). Pratical aspects of the validation of toxicity test procedures. The report and recommendations of ECVAM workshop 5. ATLA 23, 129147.[ISI]
Bennett, J. A., Pilon, V. A., Briggs, D. R., and McKneally, M. F. (1985). Evaluation of cyclosporine-treated mice as hosts for growing and testing the chemosensitivity of first-transplant-generation human tumor xenografts implanted under the kidney capsule. J. Natl. Cancer Inst. 75, 925936.[ISI][Medline]
Boardman, P. L., and Hart, F. D. (1967). Side-effects of indomethacin. Ann. Rheum. Dis. 26, 127132.[ISI][Medline]
Bradley, T. R., and Metcalf, D. (1966). The growth of mouse bone marrow cells in vitro. Aus. J. Exp. Biol. Med. Sci. 44, 287300.[ISI]
Britell, J. C., Moertel, C. G., Kvols, L. K., OConnell, M. J., Rubin, J., and Schutt, A. J. (1981). Phase II trial of iv 6-thioguanine in advanced colorectal carcinoma. Cancer Treat. Rep. 65, 909910.[ISI][Medline]
Bruce, W. R., Meeker, B. E., and Valeriote, F. A. (1966). Comparison of the sensitivity of normal hematopoietic and transplanted lymphoma colony forming cells to chemotherapeutic agents administered in vivo. J. Natl. Cancer Inst. 37, 233245[ISI][Medline]
Burke, P. J., Serpick, A. A., Carbone, P. P., and Tarr, N. (1968). A clinical evaluation of dose and schedule of administration of cytosine arabinoside (NSC 63878). Cancer Res. 28, 274279.[ISI][Medline]
Chabner, B. A., and Longo, D. L. (1996). Cancer Chemotherapy and Biotherapy, 2nd ed. Lippincott-Raven, Philadelphia, PA.
Collins, J. M., Grieshaber, C. K., and Chabner, B.A. (1990). Pharmacologically guided phase I clinical trials based upon preclinical drug development. J. Natl. Cancer Inst. 82,13211326.[Abstract]
Creasey, W. A., McIntosh, L. S., Brescia, T., Odujinrin, O., Aspnes, G. T., Murray, E., and Marsh, J. C. (1976). Clinical effects and pharmacokinetics of different dosage schedules of adriamycin. Cancer Res. 36, 216221.[Abstract]
Curren, R. D., Southee, J. A., Spielmann, H., Liebsch, M., Fentem, J. H., and Balls, M. (1995). The role of prevalidation and validation in the development, validation and acceptance of alternative methods (the first report of ECVAM Task Force of Prevalidation). ATLA 23, 409470.
Davies, J. E., Dedhia, H. V., Morgade, C., Barquet, A., and Maibach, H. I. (1983). Lindane poisonings. Arch. Dermatol. 119, 142144.[Abstract]
Davis, H. L., Jr., Rochlin, D. B., Weiss, A. J., Wilson, W. L., Andrews, N. C., Madden, R., and Sedransk, N. (1974). Cytosine arabinoside (NSC 63878) toxicity and antitumor acitivity in human solid tumors. Oncology 29, 190200.[ISI][Medline]
Davis, L. E., Alberts, D. S., Plezia, P. M., Roe, D. J., and Griswold, D. P. (1988). Predictive model for plasma concentration-versus-time profiles of investigational anticancer drugs in patients. J. Natl. Cancer Inst. 80, 815819.[Abstract]
Deldar, A. (1994). Drug-induced blood disorders: Review of pathogenetic mechanisms and utilisation of bone marrow cell culture technology as an investigative approach. Curr. Topics Vet. Res. 1, 83101.
Deldar, A., and Parchment, R. E. (1997). Preclinical risk assessment for hematotoxicity: Animal models and "in vitro" systems. In Comprehensive Toxicology (G. Sipes, C. A. McQueen, and A. J. Gandolfi, Eds.), Vol. 4, pp. 303320. Pergamon Press, New York, NY.
Deldar, A., and Stevens, C. E. (1993). Development and application of in vitro models of Hematopoiesis to drug development. Toxicol. Path. 21, 231240.[ISI][Medline]
DeVita, V. T., Carbone, P. P., Owens, A. H., Jr., Gold, G. L., Krant, M. J., and Edmonson, J. (1965). Clinical trials with 1,3-bis (2-chloroethyl)-1-nitrosourea, NSC-409962. Cancer Res. 25, 18761881.[ISI][Medline]
Dexter, T. M., Allen, T. D., Lajtha, L. G., Schofield, R., and Lord, B. I. (1973). Stimulation of differentiation and proliferation of hemopoietic cells in vitro. J. Cell. Physiol. 82, 461470.[ISI][Medline]
Dexter, T. M., and Spooncer, E. (1987). Growth and differentiation in the hemopoietic system. Ann. Rev. Cell. Biol. 3, 423441.[CrossRef][ISI]
Dexter, T. M., and Testa, N. G. (1976). Differentiation and proliferation of hemopoietic cells in culture. Meth. Cell. Biol. 14, 387405.[Medline]
Dorr, R. T., and Von Hoff, D. D. (1994). Cancer Chemotherapy Handbook, 2nd ed. Appleton and Lange, Norwalk, CT.
Du, D. L., Volpe, D. A., Grieshaber, C. K., and Murphy, M. J., Jr. (1990). Effects of L-phenylalaninemustard and L-buthionine sulfoximine on murine and human hematopoietic cells in vitro. Cancer Res. 50, 40384043.[Abstract]
Erickson-Miller, C. L., May, R., Tomaszewski, J., Osborn, B., Murphy, M. J., Jr., Page, J. G., and Parchment, R. E. (1997). Differential toxicity of Camptothecin, Topotecan and 9-Aminocamptothecin to human, canine, and murine myeloid progenitors (CFU-GM) "in vitro." Cancer Chemother. Pharmacol. 39, 467472.[CrossRef][ISI][Medline]
Fletcher, C. V., Englund, J. A., Bean, B., Chinnock, B., Brundage, D. M., and Balfour, H. H., Jr. (1989). Continuous infusion of high-dose acyclovir for serious herpesvirus infections. Antimicrob. Agents Chemother. 33, 13751378.[ISI][Medline]
Gallicchio, V. S., and Hughes, N. K. (1992). Suppression of murine hematopoiesis in vivo after chronic administration of zidovudine: Evidence that zidovudine-induced anemia is the result of decreased bone marrow-derived, erythropoietin-responsive progenitor cells. Proc. Soc. Exp. Biol. Med. 199, 459465.[Abstract]
Galton, D. A. G., and Till, M. (1955). Myleran in chronic myeloid leukaemia. Lancet 1, 425430.[ISI]
Gluckman, E., Devergie, A., Lokiec, F., Poirier, O., and Baumelou, A. (1981). Nephrotoxicity of cyclosporin A in bone-marrow transplantation. Lancet 2,144145.
Gottlieb, J. A., Mendelson, D., and Serpick, A. A. (1970). An evaluation of large intermittent intravenous doses of cyclophosphamide (NSC-26271) in the treatment of metastatic malignant melanoma. Cancer Chemother. Rep. 54, 365367.[ISI][Medline]
Grande, T., and Bueren, J. A. (1995). Analysis of the hematopoiesis in mice irradiated with 500mGy of X-rays at different stages of development. Radiation Res. 143, 327333.[ISI][Medline]
Greenberg P., Bax I., Mara, B., and Schrier S. (1974). Alterations of granulopoiesis following chemotherapy. Blood 44, 375383.[ISI][Medline]
Grem, J. L., Tutsch, K. D., Simon, K. J., Alberti, D. B., Willson, J. K., Tormey, D. C., Swaminathan, S., and Trump, D. L. (1987). Phase I study of taxol administered as a short i.v. infusion daily for 5 days. Cancer Treat. Rep. 71, 11791184.[ISI][Medline]
Gribaldo, L., Bueren, J., Deldar, A., Okland, P., Meredith, C., Moneta, D., Mosesso, P., Parchment, R., Parent-Massin, D., Pessina, A., et al. (1996). The use of "in vitro" systems for evaluating haematotoxicity. The report of recommendations of ECVAM workshop 14. ATLA 24, 211231.[ISI]
Gribaldo, L., Piccirillo, M., Casati, S., Collotta, A., Mineo, E., and Pessina, A. (1998). Drug sensitivity of Granulocyte-Macrophage precursors (GM-CFU) from fresh murine bone marrow and from long-term bone marrow cultures. Toxicol. in Vitro 12, 3945.[CrossRef][ISI]
Hamburger, A. W., Zsebo, K., and Chow, F. P. (1993). Failure of stem cell factor to ameliorate AZT-induced anemia in immunodeficient mice. Eur. J. Haematol. 50, 172178.[ISI][Medline]
Hutton, J. J., Von Hoff, D. D., Kuhn, J., Phillips, J., Hersh, M., and Clark, G. (1984). Phase I clinical investigation of 9-beta-D-arabinofuranosyl-2-fluoroadenine 5'-monophosphate (NSC 312887), a new purine antimetabolite. Cancer Res. 44, 41834186.[Abstract]
Intramural Toxicology Branch (1984). Report # 84-005. DCT, U.S. National Cancer Institute, Bethesda, MD.
Kruter, F., Eisenberger, M., Sinibaldi, V., Engstrom, C., Jacobs, M., Abrams, J., Belani, C., Gray, W., and Krasnaw, S. (1992). Phase II trial of 5 day continuous intravenous infusion of 6-thioguanine in patients with recurrent and metastatic squamous cell carcinoma of the head and neck. Invest. New Drugs 10, 8991.[ISI][Medline]
Lewis, I. D., Rawling, T., Dyson, P. G., Haylock, D. N., Juttner, D. N., and To, L.B. (1996). Standardization of the CFU-GM assay using hematopoietic growth factors. J Hematother. 5, 625630.[Medline]
Lokich, J. J., and Curt, G. (1985). A phase I and pharmacology study of continuous-infusion low-dose methotrexate administration. Cancer 56, 23912394.[ISI][Medline]
Macbeth, F. R. (1982). VM26: Phase I and II studies. Cancer Chemother. Pharmacol. 7, 8791.[ISI][Medline]
Mandel, A., and Gross, M. (1968). Agranulocytosis and phenothiazines. Dis. Nerv. Syst. 29, 3236.[ISI][Medline]
Manouilov, K. K., White, C. A., Boudinot, F. D., Fedorov, I. I., and Chu, C. K. (1995). Lymphatic distribution of 3'-azido-3'-deoxythymidine and 3'-azido-2', 3'-dideoxyuridine in mice. Drug Metab. Dispos. 23, 655658.[Abstract]
Metcalf, D. (1984). The basic biology of hematopoiesis. In The Haematopoietic Colony Stimulating Factors (D. Metcalf, Ed.), pp. 126. Elsevier, Amsterdam.
Moertel, C. G., Schutt, A. J., Reitemeier, R. J., and Hahn, R. G. (1972). A comparison of 5-fluorouracil administered by slow infusion and rapid injection. Cancer Res. 32, 27172719.[ISI][Medline]
Naughton, D. A., Sibanda, D., Azar, L., and San Roman, J. (1992). Differential effects of drugs upon hematopoiesis can be assessed in long-term bone marrow cultures established on nylon screens. Proc. Soc. Exp. Biol. Med. 199, 481490.[Abstract]
Neelis, K. G., Dubbelmann, J. D., Luo, Q. L., Thomas, G. R., Eaton, D. L., and Wagemaker, G. (1997). Simultaneous administration of TPO and G-CSF after cytoreductive treatment of Rhesus monkeys prevent thrombocytopenia, accelerates platelet and red cell reconstitution, alleviates neutropenia, and promotes the recovery of immature bone marrow cells. Exp. Hematol. 25, 10841093.[ISI][Medline]
Nissen, N. I., Larsen, V., Pedersen, H., and Thomsen, K. (1972). Phase I clinical trial of a new antitumor agent, 4'-demethylepipodophyllotoxin 9-(4,6-O-ethylidene-beta-D-glucopyranoside) (NSC-141540; VP-16213). Cancer Chemother. Rep. 56, 769777.[ISI][Medline]
Noble, C., and Sina, J. F. (1993). Usefulness of the in vitro bone marrow colony forming assay in cellular toxicology. In Vitro Toxicol. 6, 187195.
Parchment, R. E. (2000). Oncology clinical trials and in vitro prediction models for xenobiotic toxicology. Develop. Animal Vet. Sci. 31, 693708.
Parchment, R. E., Gordon, M., Grieshaber, C. K., Sessa, C., Volpe, D., and Ghielmini, M. (1998). Predicting hematological toxicity (myelosuppression) of cytotoxic drug therapy from in vitro tests. Ann. Oncol. 9, 357364.[Abstract]
Parchment, R. E., Huang, M., and Erickson-Miller, C. L. (1993). Roles for "in vitro" myelotoxicity tests in preclinical drug development and clinical trial planning. Toxicol. Pathol. 21, 241250.[ISI][Medline]
Parchment, R. E., and Murphy, M. J., Jr. (1997). Human hematopoietic stem cells: Laboratory assessment and response to toxic injury. In Comprehensive Toxicology (G. Sipes, C. A. McQueen, and A. J. Gandolfi, Eds.), Vol. 4, pp. 335361. Pergamon Press, New York, NY.
Parchment, R. E., Volpe, D. A., LoRusso, P. M., Erickson-Miller, C. L., Huang, M., Murphy, M. J., Jr., and Grieshaber, C. K. (1994). An in vivo-in vitro corrleation of the myelotoxicity of 9-methoxypyrazoloacridine (PZA, NSC-366140, PD115934) to myeloid and erythroid hematopoietic progenitors from human, murine, and canine marrow. J. NatI. Cancer Inst. 86, 273280.[ISI]
Parent-Massin, D., Guern, G., Leglise, C., Sensebe, L., Berthou, C., Abgrall, J. F., and Riche, C. (1993). Revelance of in vitro studies in drug-induced agranulocytosis. Report of 14 cases. Drug Safety 9, 461469.
Parent-Massin, D., and Parchment, R. E. (1998). Haematotoxicity of mycotoxins. Rev. Med. Vet. 149, 591598.[ISI]
Parent-Massin, D., and Thouvenot, D. (1995). In vitro toxicity of trichothecenes on rat haematopoietic progrenitors. Food Addit. Contam. 12(1), 4149.[ISI][Medline]
Pessina, A. (1998). The granulocyte macrophage colony-forming unit assay. In Animal Cell Culture Techniques (M. Clynes, Ed.), pp. 217230. Springer-Verlag, Berlin.
Pessina, A., Albella, B., Bueren, J., Brantom, P., Casati, S., Gribaldo, L., Croera, C., Gagliardi, G., Foti, P., Parchment, R., et al. (2001). Prevalidation of a model for predicting acute neutropenia by colony forming unit granulocyte/macrophage (CFU-GM) assay. Toxicol. in Vitro 15, 729740.[CrossRef][ISI][Medline]
Pessina, A., Piccirillo, M., Mineo, E., Catalani, P., Gribaldo, L., Marafante, E., Neri, M. G., and Raimondi, A. (1999). Role of SR-4987 stromal cells in the modulation of doxorubicin toxicity to in vitro granulocyte-macrophage progenitors (CFU-GM). Life Sci. 65, 513523.[CrossRef][ISI][Medline]
Petrakis, N. L., Bierman, H. R., Kelly, K. H., White, L. P., and Shimkin, M. B. (1954). The effect of 1,4-dimethanesulfonyloxybutane (GT-41 or Myleran) upon leukemia. Cancer 7, 383390.[ISI]
Pizzo, P. A., Eddy, J., Falloon, J., Balis, F. M., Murphy, R. F., Moss, H., Wolters, P., Brouwers, P., Jarosinski, P., Rubin, M., et al. (1988). Effect of continuous intravenous infusion of zidovudine (AZT) in children with symptomatic HIV infection [see comments]. N. Engl. J. Med. 319, 889896.[Abstract]
Rivera-Calimlim, L. (1982). Problems in therapeutic blood monitoring of chlorpromazine. Ther. Drug Monit. 4, 4149.[ISI][Medline]
Robinson, M. J., Nichols, E. A., and Taitz, L. (1967). Nephrotoxic effect of acute sodium salicylate intoxication in the rat. Arch. Pathol. 84, 224226.[ISI][Medline]
San Roman, J., Kamali, V., Sibanda, B., Weintraub, J. P., Gee, J. M., and Naughton, B. A. (1994). Measurement of differential hematotoxicity using long-term bone marrow cultures. In Vitro Toxicol. 7, 291299.
Schoeters, G. E. R., Van der Plaetse, F., Leppens, H., Van Vlasselaer, P., and Van Den Heuvel, R. (1995). Haemopoietic and osteogenic toxicity testing in vitro using murine bone marrow cultures. Toxicol. in Vitro 9(4), 421428.[CrossRef][ISI]
Schweikart, K. (2000). Toxicology and Pharmacology Branch, U.S. National Cancer Institute, Washington, DC.
Seifert, P., Baker, L. H., Reed, M. L., and Vaitkevicius, V. K. (1975). Comparison of continuously infused 5-fluorouracil with bolus injection in treatment of patients with colorectal adenocarcinoma. Cancer 36, 123128.[ISI][Medline]
Southern Research Institute (1980). Toxicity testing of anticancer drugs in small experimental animals 5/16/608/23/80. Toxicology and Preclinical Pharmacology Branch, U.S. National Cancer Institute, Birmingham, AL.
Stahelin, H. (1970). 4'-Demethyl-epipodophyllotoxin thenylidene glucoside (VM 26), a podophyllum compound with a new mechanism of action. Eur. J. Cancer 6, 303311.[ISI][Medline]
Starr, H. G., Jr., and Clifford, N. J. (1972). Acute lindane intoxication: A case study. Arch. Environ. Health 25, 374375.[ISI][Medline]
Van Den Heuvel, R. L., Schoeters, G., and Leppens, H. (1997). Haematotoxicity testing in vitro using human cord blood haemopoietic cells. Toxicol. in Vitro 11, 689693.[CrossRef][ISI]
Veerman, G., Van Haperen, R., and Vermorken, J. B. (1994). Superior therapeutic activity of prolonged compared with bolus administration of 2'2' difluorodeoxycytidine (gemcitabine) in vivo against murine tumors. In Biochemical Pharmacology of Gemcitabine (2'2' difluorodeoxycytidine) (R. van Haperen, Ed.), 9 vols, pp. 189206. University Press, Amsterdam.
Vogelzang, N. J., Ruane, M., and DeMeester, T. R. (1985). Phase I trial of an implanted battery-powered, programmable drug delivery system for continuous doxorubicin administration. J. Clin. Oncol. 3, 407414.[Abstract]
Volpe, D. A., Tomaszewski, J. E., Parchment, R. E., Garg, A., Flora, K. P., Murphy, M. J., Jr., and Grieshaber, C. K. (1996). Myelotoxic effects of the bifunctional alkylating agent Bizelesin to human, canine, and murine myeloid progenitor cells. Cancer Chemother. Pharmacol. 39, 143149.[CrossRef][ISI][Medline]
Yeh, K. C. (1985). Pharmacokinetic overview of indomethacin and sustained-release indomethacin. Am. J. Med. 79, 312.