* National Center for Environmental Assessment, Office of Research and Development, U.S. Environmental Protection Agency, Washington, D.C. 20460;
Department of Biostatistics, Harvard School of Public Health, Boston, Massachusetts 02115; and
Pathology Associates, International, Frederick, Maryland 21701
Received March 14, 2002; accepted June 25, 2002
![]() |
ABSTRACT |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Key Words: Habers Law; hyperthermia; whole embryo culture; developmental toxicity; risk assessment; rat; dose-rate; dose-duration; heat shock.
![]() |
INTRODUCTION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
In order to study the interaction of exposure concentration (or level) and duration, rat embryos were exposed in whole embryo culture to increased temperature and evaluated for alterations of development 24 h later. The use of hyperthermia and whole embryo culture was chosen for several reasons. First, our laboratory has substantial experience in studying the effects of hyperthermia on development. Exposure of gestation day (GD) 10 rat embryos to elevated temperatures, either in utero or in whole embryo culture, results in altered development of many growth and morphological parameters (Cuff et al., 1993; Kimmel et al., 1993a
,b
). From the whole embryo culture studies, it was noted that the pattern of response of virtually all of the evaluated parameters was defined by both the temperature level and duration of exposure (Kimmel et al., 1993b
). Second, heat is a physical agent that is not influenced by metabolic factors, that can be carefully controlled in whole embryo culture, and that affects development following brief exposures. Finally, embryos in culture are exposed individually rather than via the dam, which decreases the need to account for intralitter correlation of responses (Williams et al., 1996
). All of these factors simplify modeling the response to exposure of a biological entity (the embryo), which is constantly changing as development proceeds.
The specific purpose of the study was to assess whether a response to hyperthermia depends solely on the multiple of temperature and duration (i.e., whether Habers Law applies), or whether there is an additional component of exposure needed to model the relationship. The models developed by Scharfstein and Williams (1994) were used to assess the compatibility with and/or departure of the responses to hyperthermia from Habers Law. Response-surface models were developed as a useful approach for evaluating the combined effect of more than one variable on a variety of responses through model-fitting and 3-dimensional graphs. Eventually, the goal is to develop models that permit extrapolation of data to untested exposure intensities and durations and apply these in the risk assessment process.
![]() |
MATERIALS AND METHODS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Embryos were isolated from the dams on GD 10 and Reicherts membrane was opened to expose the visceral yolk sac. Embryonic heart beat, yolk sac circulation, flexion, and position of the neural folds were noted as general signs of viability and developmental stage. Embryos used for culture were at or near the stage of neural fold apposition (1012 somites in our laboratory). The only exception to the previous culture technique was that only one embryo was incubated per vial.
Embryos were equilibrated at 37°C for approximately 2 h before exposure to one of the temperature-duration combinations. Exposure to heat was carried out in a water bath (Precision 181, Cole-Palmer), maintained within a 0.5°C temperature range (e.g., 40.0°C = 40.040.5; 40.5°C = 40.541.0; etc.). A total of 432 embryos were cultured for this study. Table 1 shows the 6 x 7 (temperature x duration) design employed in this study, and the number of embryos cultured at each temperature-duration combination. Following heat exposure, the vials were equilibrated to 37°C in a water bath and returned to the 3738°C incubator/rotator for the remainder of the 2022 h incubation. Preliminary studies indicated that both attaining the increased temperature and returning to 37°C after exposure were accomplished in less than 30 s.
|
The effect of the hyperthermia on normal development was defined for both individual growth and morphological parameters, as well as for viability and collective measures of alteration. A nonviable embryo was any embryo without a heart beat and yolk sac circulation at the time of evaluation. If the embryo exhibited even a faint heart beat or yolk sac circulation, it was considered viable and evaluated for alterations in growth and morphology. For those embryos which were judged to be viable, a collective measure of whether the embryo was altered was developed, using the following criteria: yolk sac diameter, < 3.2 mm; crown rump length, < 2.9 mm; head length, < 1.4 mm; proboscis length, < 0.5 mm; somite number, < 19; somite integrity, either abnormal structure or segmentation. Somite number and integrity were evaluated on both sides of the embryo. An effect on either side was considered an alteration in normal development. Any one or more of these 6 criteria or a mean of the 13 morphological parameters < 2.69 could be the basis for considering an embryo altered. A collective measure of affected embryos included any embryo that was either nonviable or altered.
Several statistical procedures were applied to evaluate the effects of hyperthermia. First, the effects of temperature and duration of exposure were assessed on each individual outcome in separate univariate models. For continuous outcomes (head length, crown rump length, yolk sac diameter, number of left and right somites, and proboscis length), a general linear regression model for predicting the mean response as a function of the exposure covariates was employed. For the morphological parameters that were assessed on an ordinal scale (1 = severe, 2 = moderate, 3 = normal), an ordinal logistic regression model was fit to each outcome to assess the effect of exposure covariates on the probability of response within each severity level. The MMS was treated as a continuous variable, and a linear regression model was also used for this response. The binary responses (nonviable, altered, and affected) were modeled via standard logistic regression models as described below.
A common approach for incorporating duration of exposure and temperature was utilized in all three types of models, i.e., the linear regression models for continuous outcomes, ordinal logistic regression models for ordinal outcomes of morphological severity, and standard logistic regression models for viability, altered and affected status. Ignoring duration of exposure, the usual dose-response model for a binary outcome models the probability p(d) of an adverse outcome as a function of dose level, as follows:
![]() | (1) |
where F represents the link function, such as a logistic or probit link function. In the context of the hyperthermia experiments, the dose level d represents the increase in temperature over the control temperature of 37°C. Scharfstein and Williams (1994) describe how the above model can be extended to include a covariate that represents the cumulative exposure, d*t, along with an additional covariate t, which reflects the extra effect of duration after controlling for cumulative exposure:
![]() | (2) |
In the current study, the duration of exposure at the control temperature was assumed to have no adverse effect on development, so the model above was modified such that the duration effect, t, would only be included for temperatures greater than 37°C:
![]() | (3) |
where t = 1 if d > 37 and 0 otherwise. If Habers Law of equal responses for any constant cumulative exposure holds, then Model 3 can be reduced to:
![]() | (4) |
In other words, the implication of Habers Law is that ß1 = 0. By fitting Model 3, one can conduct a statistical test of the significance of ß1 to assess whether Habers law is consistent with the observed data. In applying Models 3 and 4 to the hyperthermia data, a logistic link was employed.
For the ordinally scaled morphological outcomes, Models 3 and 4 were extended by estimating Pr(Y r | d, t) using standard ordinal logistic regression models, with r = 1, 2, or 3 for severely affected, moderately affected, and normal, respectively. This approach for incorporating duration of exposure into dose-response models can also be applied for the continuous outcomes. In this situation, the mean response is modeled as a linear function of the same covariates, as follows:
![]() | (5) |
where the coefficient ß1 represents the effect of duration accounting for cumulative exposure. For example, the mean µ(d, t) might reflect the mean crown-rump length as a function of duration (t) and temperature level (d).
The results of fitting these models to the hyperthermia data are illustrated in two ways: (1) the p values for tests of significance of the effects of cumulative exposure (d*t) and duration (t) are provided from the univariate models for the growth outcomes, and for the binary outcomes of nonviable, altered, and affected embryos; and (2) 3-dimensional response-surface plots are generated that exhibit the actual or predicted response levels at combinations of duration of exposure and temperature level. The levels of temperature and duration of exposure form a horizontal plane, while the response level forms a surface over this plane.
The response-surface plot is very useful for visualizing the joint effects of temperature and duration of exposure. Their use originated in engineering applications aimed at maximizing the yield of some output, but more recently they have been applied to a variety of other contexts in both design and analysis of experimental data (Box and Draper, 1987). Several different types of response-surface plots are provided. The first illustrates the actual experimental means for a given response, represented as a pyramid at a coordinate determined by the duration of exposure and temperature and a height reflecting the response level (example, Fig. 1a
). This plot gives an indication of the inherent variability in response data and of the gaps in areas where data was not collected. The second type of plot is a smoothed spline interpolation applied to the points on the response-surface plot (example, Fig. 1b
), which filters out some of the random variability by taking weighted averages over all response points to determine the smoothed response-surface levels. Finally, there are two plots that illustrate the predicted response levels using the modeling techniques described above. Figure 2a
illustrates the predicted response surface after enforcing Habers Law by applying Model 4. Figure 2b
shows the predicted response surface using the general, or extended model given by Model 3.
|
|
![]() |
RESULTS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
|
|
Summarizing the combined effects, the model that best fits the proportion of animals showing some effect includes a parameter that is in addition to the constant multiple of temperature and duration. This also appears to be the case for nonviable and altered embryos. Moreover, when the multiple of temperature and duration is held constant, short exposures to higher temperatures result in a greater proportion of affected embryos than longer exposures to lower temperatures.
Growth Parameters
Table 5 summarizes the response of 6 growth parameters as they relate to the covariates, i.e, the cumulative exposure effect (d*t), and the extra effect of duration after controlling for cumulative exposure (t). The values were obtained from fitting Model 5. With the exception of yolk sac diameter, all of the individual growth parameters were sensitive to changes in temperature and duration of exposure. The fact that both d*t and t were consistently significant (p < 0.05) indicates that Habers Law did not adequately describe the effects of hyperthermia on individual growth parameters. In particular, since the coefficient on the duration effect (ß1) was consistently estimated to be positive for each of the growth parameter models, this suggests that short acute exposures result in more retardation of growth parameters than do longer lower-level exposures within the same d*t multiple. Although significant effects of both exposure variables were observed, the proportion of the total variability explained by these covariates was relatively low; the R-squared values ranged from 811%. This implies that while both exposure covariates have a significant impact on the mean response for each growth parameter (except yolk sac diameter), there still remains considerable variability around the mean in the responses of embryos exposed to the same exposure level and duration.
|
|
|
![]() |
DISCUSSION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
The current study demonstrates that the response of the developing embryo to hyperthermia is dependent on an additional component of exposure duration beyond the cumulative exposure. This is true for the combined developmental effects as well as for all but three of the individual endpoints evaluated. Thus, for the vast majority of parameters measured in this study, Habers Law was not sufficient by itself to describe the relationship of responses to various temperature-duration combinations. Moreover, the probability of an effect was greater at higher temperatures for shorter durations than at lower temperatures for longer durations, given the same temperature-duration multiple, and this pattern of response is generally consistent across endpoints.
The temperature and duration of hyperthermia at which effects were first observed in the current study are below those levels that have been routinely used in studying the effects of hyperthermia on embryonic development. Generally, both in vivo and in vitro, exposure temperatures or core body temperatures of 42°C or higher have been employed for periods exceeding 15 min (e.g., Mirkes, 1985; Shiota, 1988
; Walsh et al., 1987
). Mirkes (1985) did report a slight effect at 41°C in vitro, but it required more than 2 h of exposure. In an in vivo study that more closely approximates the current study design, Germaine et al.(1985) maintained GD 9 rats at various core body temperatures (ranging from 40.5°C to 43.5°C) for various durations (a "spike" or from 2 min to 480 min). They observed litters with abnormal fetuses at all but the lowest core temperature (40.5°C). At 40.5°C, they did not observe an effect, even after 8 h of exposure. In the present study, effects were observed at temperatures as low as 40.0°C, following approximately 3060 min of exposure.
Although it is not possible to directly compare our in vitro results with the in vivo study of Germaine et al.(1985), the lowest temperatures (4041°C) used in the two studies are within the range of potential human exposures. Hyperthermia in the human can result from febrile illness, as well as from occupational and environmental conditions; and associations of hyperthermia with developmental alterations in humans have been made (Edwards et al., 1995; Graham et al., 1998
). Under in vivo conditions in our laboratory, the GD 10 maternal Sprague-Dawley rat maintains a core body temperature of 3738°C (Kimmel et al., 1993a
). Thus, our results suggest that even at an elevation of only 23°C above core body temperature, the duration required to achieve an effect may be less than 60 min. This is similar to previous reports of threshold elevations of 1.52.5°C above the core body temperature being required for alterations in development (Edwards, 1986
; Graham et al., 1998
).
It should be noted that the responses observed at 37°C can be considered normal variation within the test system. Embryos that equilibrate and are maintained in culture over 2224 h at 37°C generally develop comparable to embryos in vivo. Occasionally, even an unexposed embryo will show an observable alteration in development. Another factor that may result in an affected embryo in the 37°C group is a misclassification during the morphological evaluation. The number of embryos per group, while large by some standards, will still result in an 810% increase for every affected embryo. The way the model is defined, we deliberately do not take into account duration of exposure at 37°C. However, this does not imply that we assume there to be no baseline effect. Rather, we assume the baseline effect to be constant over duration of exposure for the 37°C heat level. In fitting the model, an overall average response was estimated for all of those embryos exposed to 37°C, and the response predicted by the model would be the same for all embryos in this exposure group.
The in vitro induction of hyperthermia, as used in this study, provides an ideal system for the initial modeling of the relationship of level and duration of exposure on embryonic development. Heat is a physical agent that appears to act directly on the embryo (Cuff et al., 1993; Kimmel et al., 1993b
). This, coupled with the isolation of the embryo in culture, removes potential confounders such as maternal tissue layers, and metabolic and pharmacokinetic factors, that may modulate the observed response. Our previous studies have shown that embryonic development in whole embryo culture is comparable to what is observed in vivo, both in control and heat-treated embryos, at least over the 24-h period used in this study (Cuff et al., 1993
; Kimmel et al., 1993b
). The whole embryo culture also permits the level of the exposure to be controlled within clearly defined ranges (0.5°C in this study) for discrete periods of time. Finally, the duration of exposure required to obtain an effect is relatively short, effectively removing the influence that normal developmental changes may have on the susceptibility of a particular endpoint.
If the statistical model employed in this study is to be of general use for understanding exposure-duration relationships, it will have to be shown to be more generally applicable to chemical and other physical agents. Weller et al.(1999) have recently reported a similar pattern of response for many of the same developmental parameters in an in vivo study of ethylene oxide exposure during pregnancy in mice. In that study, high exposure levels of shorter duration had a much greater effect on fetal viability, growth, and malformations than low exposure levels at longer durations, given a constant concentration-duration multiple. Other investigators have also reported divergence from Habers Law for neurotoxic effects (Bushnell, 1997; Crofton et al., 1996
). The responses were dependent on dose-rate, but not in a strict C x T relationship.
The development of appropriate models should more clearly define the relationship of exposure concentration and duration. Initial application of these models will require broad assumptions about the mode of action. Ultimately, as our understanding of operative modes of developmental toxicity increases, such models should permit interpolation of data to untested C x T combinations, reducing the uncertainty in assessing the data and increasing our confidence in predicting potential human risk. The current study suggests that Habers Law is likely to underestimate the probability of an effect when extrapolating to higher exposure levels for shorter durations and overestimate the probability at lower exposure levels for longer durations. The more restrictive Habers model (Model 4) underestimates the probability of an abnormal embryo by as much as 30% at the shorter exposure durations and high temperatures (e.g., 510 min at 42°C), while it overestimates the probability of an abnormal embryo by 2030% at the longer durations at low temperatures (e.g., 3060 min at 40.541°C). In contrast, the extended model (Model 3) provides predicted probabilities that are closer to those actually observed. This may vary depending on the relative exposure period, the endpoint being investigated, and the proximity to the target site. It should be reemphasized that Habers Law was originally developed to apply to a specific exposure scenario based on specific assumptions. There is no reason to expect it to fit every situation, especially when being applied to the developing embryo, which presents a constantly changing exposure target.
Current practice in risk assessment is to use Habers Law for duration adjustment when extrapolating from shorter to longer exposure durations, but not when adjusting from longer to shorter exposure periods. This appears to be a conservative approach based on the data from this study and those by other investigators (Bushnell, 1997; Weller et al., 1999
). In general, no duration adjustment is currently made for developmental toxicity data, based on the assumption that a single, discrete exposure during a critical developmental period is all that is necessary to cause an alteration (U.S. EPA, 1991
). However, the results of modeling in this study and that of Weller et al.(1999) show a significant effect of exposure-duration on developmental responses and suggest that continuing this practice may not be conservative enough to adequately protect against potential developmental toxicity when predicting responses from shorter to longer durations. Further efforts to describe exposure-duration relationships will foster development of more predictive models for use in risk assessment for developmental toxicity.
![]() |
ACKNOWLEDGMENTS |
---|
![]() |
NOTES |
---|
This work was supported in part by a grant from the National Institute of Environmental Health Science. The views expressed herein are those of the authors and not necessarily the views of the U.S. Environmental Protection Agency.
Portions of this data were presented at the U.S. EPA Workshop on the Relationship Between Exposure Duration and Toxicity, August 56, 1998, Washington, DC.
![]() |
REFERENCES |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Bushnell, P. J. (1997). Concentration-time relationships for the effects of inhaled trichlorethylene on signal behavior in rats. Fundam. Appl. Toxicol. 36, 3038.[ISI][Medline]
Crofton, K. M., Padilla, S., Tilson, H. A., Anthony, D. C., Raymer, J. H., and MacPhail, R. C. (1996). The impact of dose rate on the neurotoxicity of acrylamide: The interaction of administered dose, target tissue concentrations, tissue damage, and functional effects. Toxicol. Appl. Pharmacol. 139, 163176.[ISI][Medline]
Cuff, J. M., Kimmel, G. L., Kimmel, C. A., Heredia, D. J., Tudor, N., and Chen, J. (1993). Relationship between abnormal somite development and axial skeletal defects in rats following heat exposure. Teratology 48, 259266.[ISI][Medline]
Eastern Research Group (1998). Summary of the U.S. EPA Workshop on the Relationship Between Exposure Duration and Toxicity. Prepared for the U. S. Environmental Protection Agency, Risk Assessment Forum, Washington, D.C., under Contract No. 68-D50028. Available from: Technical Information Staff, (202-564-3261).
Edwards, M. J. (1986). Hyperthermia as a teratogen: A review of experimental studies and their clinical significance. Teratog. Carcinog. Mutagen. 6, 563582.[ISI][Medline]
Edwards, M. J., Shiota, K., Smith, M. S., and Walsh, D. A. (1995). Hyperthermia and birth defects. Reprod. Toxicol. 9, 411425.[ISI][Medline]
Germaine, M.-A., Webster, W. S., and Edwards, M. J. (1985). Hyperthermia as a teratogen: Parameters determining hyperthermia-induced head defects in the rat. Teratology 31, 265272.[ISI][Medline]
Graham, J. M., Jr., Edwards, M. J., and Edwards, M. J. (1998). Teratogen update: Gestational effects of maternal hyperthermia due to febrile illnesses and resultant patterns of defects in humans. Teratology 58, 209221.[ISI][Medline]
Jarabek, A. M. (1995). Consideration of temporal toxicity challenges current default assumptions. Inhalation Toxicol. 7, 927946.[ISI]
Kimmel, C. A., Cuff, J. M., Kimmel, G. L., Heredia, D. J., Tudor, N., Silverman, P. M., and Chen, J. (1993a). Skeletal development following heat exposure in the rat. Teratology 47, 229242.[ISI][Medline]
Kimmel, G. L. (1995). Exposure-duration relationships: The risk assessment process for health effects other than cancer. Inhalation Toxicol. 7, 873880.[ISI]
Kimmel, G. L., Cuff, J. M., Kimmel, C. A., Heredia, D. J., Tudor, N., and Silverman, P. M. (1993b). Embryonic development in vitro following short-duration exposure to heat. Teratology 47, 243251.[ISI][Medline]
Mirkes, P. E. (1985). Effects of acute exposures to elevated temperatures on rat embryo growth and development in vitro. Teratology 32, 259266.[ISI][Medline]
Pierano, W. B., Mattie, D., and Smith, P. (1995). Proceedings of the conference on temporal aspects in risk assessment for noncancer endpoints. Inhalation Toxicol. 7, 8371029.[ISI]
Scharfstein, D. O., and Williams, P. L. (1994). Design of developmental toxicity studies for assessing joint effects of dose and duration. Risk Anal. 14, 1057107.[ISI][Medline]
Shiota, K. (1988). Induction of neural tube defects and skeletal malformations in mice following brief hyperthermia in utero. Biol. Neonate 53, 8697.[ISI][Medline]
U.S. EPA (1991). Guidelines for developmental toxicity risk assessment. Fed. Regist. 56, 6379863826.
Walsh, D. A., Klein, N. W., Hightower, L. E., and Edwards, M. J. (1987). Heat shock and thermotolerance during early rat embryo development. Teratology 36, 181191.[ISI][Medline]
Weller, E., Long, N., Smith, A., Williams, P., Ravi, S., Gill, J., Henessey, R., Skornik, W., Brain, J., Kimmel, C., Kimmel, G., Holmes, L., and Ryan, L. (1999). Dose-rate effects of ethylene oxide exposure on developmental toxicity. Toxicol. Sci. 50, 259270.[Abstract]
Williams, P. L., Molenbergs, G., and Lipsitz, S. (1996). Analysis of multiple ordinal outcomes in developmental toxicity studies. J. Agricultural Biol. Environ. Stat. 1, 125.
Witschi, H. (1999). Some notes on the history of Habers Law. Toxicol. Sci. 50, 164168.
Witschi, H. (1997). The story of the man who gave us Habers Law. Inhalation Toxicol. 9, 201209.[ISI]