The most common cancers in laboratory rodents are liver cancers – both adenomas and carcinomas. There has been a long argument about the relative merits of combining them or considering them separately in the interpretation of long-term bioassays for chemical carcinogenesis. In this paper we examine various aspects of the liver adenomas and liver carcinomas as seen in the CBDS and TDMS databases of the National Toxicology Program. It appears that the data themselves demonstrate interesting differences between the behavior of these tumors which probably have biological origin. Specifically, we find a strong anticorrelation between the appearance of adenomas and carcinomas in the same animal in both control and chemically-treated groups. This relationhship that does not seem to result from differential survival but may be influenced by the animal’s body weight. We hope that toxicologists and pathologists will be encouraged to preserve the pathological distinctiveness of the two tumor types, in spite of possible difficulty in diagnosis, when analyzing rodent bioassays.

1. Introduction

Liver tumors are the most common response of rodents to chemical exposure in modern long term rodent cancer bioassays (Gold et al., 1991). In fact  a deliberate decision was made 30 years ago to choose strains of rodents with a high sensitivity of the liver to known carcinogens. These strains show a large rate of tumors in untreated animals.   They play a major role in the determination of carcinogenicity for many chemicals with widespread exposure potential (IRIS 1999).

The primary chemical-induced tumors in the rodent liver are benign hepatocellular adenomas and malignant hepatocellular carcinomas (Gold et al., 1991). It is common practice in the evaluation of bioassays to lump these two tumor types in statistical evaluation of dose-response (McConnell et al., 1986). Combining of benign adenomas and malignant carcinomas often by the fact that the sensitivity to detect weak carcinogens is thereby enhanced. Such combining is justified scientifically if liver adenomas are precursor lesions to carcinomas, as suggested by (Ward et al., 1996). However, others suggest that liver adenomas and carcinomas are biologically distinct tumor types and combining them may be inappropriate (Butler, 1996). For example, there is evidence that, in contrast to carcinomas, benign tumors of the rat liver may regress after the cessation of chemical treatment (McConnell et al.1986, Butler, 1996). Some analysts argue, therefore that only the malignant carcinomas should be used in calculations of a risk of a chemical exposure. However Byrd et al. (1990) showed that mouse liver adenomas were equally predictive of any rat tumor as liver carcinomas. Previous analyses from this group (e.g, Linkov et al, 1998a, 1998b; Gray et al., 1995, Byrd et al., 1990) have placed liver adenomas and carcinomas in distinct categories. Criticism of this approach (J. Haseman, personal communication) prompted this study to examine the correlation between adenomas and carcinomas in rodents and their relationship with animal body weight, an attribute known to influence liver tumor rates (Hart et al., 1995a; Hart et al., 1995b; Keenan et al., 1996; Turturro et al., 1996). In an earlier paper (Linkov et al., submitted) we show that the anticorrelation noticed by Young and Gries (1984) and by Haseman et al. (1997) between all liver tumors combined and all lymphomas combined is different when liver carcinomas and liver adenomas are kept separate. In this paper we find strong evidence that adenomas and carcinomas have different biological antecedents and effects. We also examine the impact of combining, or not, on qualitative assessment of carcinogenic potential of chemicals.?

2. Methods

2.1 Database

Two data bases are used for this study.  The first is the data on tumor rates in rodents collected in the Carcinogenesis Bioassay Data System (CBDS) which report the results of rodent bioassays conducted by the National Toxicology Program (NTP) (initially in the National Cancer Institute (NCI) and later in the National Institute for Environmental Health Sciences (NIEHS from about 1970 to 1983 and the second is the Toxicology Database Management System (TDMS) used after 1985.  The bioassays that are used in this study include a concurrent control group, last 70 weeks or more, and have been performed with both sexes in 2 species (rats and mice).  For the CBDS data base these criteria result in bioassays for 312 chemicals.   For the TDMS data base bioassays for 139 chemicals were used. These were provided to us by Dr Angelo Turturro of NCTR although we now have a more complete set of individual animal records.

The analyses follow methods detailed in Byrd et al (1991), Gray et al. (1995) and Linkov, et al., (1998a, 1998b).   For analysis of animals in the control (no dose) group all untreated animals from the 312 bioassays were combined.  When examining treated animals, all animals in the low dose group from the 312 bioassays were combined and all animals in the high dose group from all the 312 bioassays were combined.

2.2  Tumor classifications

Tumors in the CBDS database  are classified according to 20,926 (SNOP) codes. We used asimilar tumor categorization system to that used in Byrd et al. (1991), Gray et al. (1995) and Linkov et al. (1998a, 1998b) was implemented in this work. The over twenty thousand SNOP codes are combined into a total of 102 classifications which include both tumor types and tumor locations (Linkov et al., 1998a - Table 1). Among these 102 classifications, 66 were classified as malignant primary neoplasms and 36 as benign primary neoplasms. Table 1 details the tumor types, and their occurrence in both control and all animals, considered as liver adenomas and carcinomas in this analysis. There is a small difference between in this work and the earlier papers mentioned above. Liver neoplastic nodules are classified with adenomas (class 7) instead of (incorrectly) among malignant carcinomas (class 64).
 
 

[[WHY ARE THERE STILL ENTRIES FOR NEOPLASTIC NODULES IN THE CARCINOMA SECTION????]]

For the TDMS data base liver adenomas and carcinomas were already grouped in the original pathology classification.
 
 

2.3 Analysis of correlations
 

Correlations were analyzed using 2X2 contingency tables as shown in Figure 1. Correlations were studied at the individual animal level. That is, for each tumor type it was noted how many individual animals had both liver adenomas and liver carcinomas (+/+); only adenomas (+/-); only carcinomas (-/+); or neither tumor (-/-).

Gart et al. (1986) recommended the odds ratio methodology for quantitative analysis of the association among tumor types. We follow their advice, but in addition present results as probabilities and conditional probabilities. Correlations were assessed in three ways: (1) through a chi-square analysis of difference from independent distribution, (2) through construction of the odds ratio ((+/+)*(-/-))/((+/-)*(-/+)) and its standard error, and (3) by calculation of the probabilities of finding each tumor and the conditional probability of finding tumor 1 given the presence of tumor 2. An example presented in Table 2.

Table 2. 2x2 Table for Correlation Between Liver Adenomas (Classification 7) and Liver Carcinomas (Classification 64) in Male Mice of the Control Group.
 
 

	Liver Adenoma
		+	-	total
Liver	+	116	2461	2557
Carcinoma	-	1242	11395	12637
	total	1358	13856	15214

The odds ratio (OR) is 0.43 with a standard error of 0.04.   The c²(chi-square) with 1 degree of freedom (df) is 74.72 leading to a p<0.01 (c² =3.84 is significant with a probability due to chance p < 0.05).

The probability of finding an adenoma P₇ is 1358/15214 = 0.09
The probability of finding a carcinoma is P₆₄ is 2557/15214 = 0.16
The probability of finding a carcinoma given an adenoma (P_64|7 = P₆₄ ^Ç ₇/P₇) is 116/2557 = 0.05
P_64|7/P₆₄ = 0.50 +/- 0.2 showing unequivocally the negative correlation.  Calculation of the uncertainty in the probabilities is easily obtained by noting that the primary uncertainty is the sampling error of the expectation for the ++ entry in the above table. The expectation for the ++ entry in the absence of a correlation is: 2557*1358/15214 = 228.

3. Results

3.1 Correlations between liver adenoma and liver carcinoma

We first examined how often individual animals had a liver adenoma, carcinoma, both or neither tumor type. Table 3(a) presents the correlation between liver adenomas and carcinomas at the individual animal level for both male and female mice from the control, low dose, and high dose groups in the bioassays. Except for control female mice, there is a strong and highly statistically significant anticorrelation between liver adenomas and carcinomas. Interestingly, in the control female mice there is a statistically significant positive correlation between liver adenomas and carcinomas.

We further investigated this comparison by dividing the animals into two groups, those that survived until the scheduled sacrifice at the end of the experiment, and those that died early. Since carcinomas are likely to be more lethal tumors, we expected that there might be a stronger negative correlation in the group of animals that died before scheduled sacrifice since adenomas would not have time to develop (if the tumor types are indeed independent). However, tables 3(b) and (c) show that the anticorrelation was stronger in the animals that survived to scheduled sacrifice. All groups of scheduled sacrifice animals have odds ratios smaller than 1 and only control female mice are not statistically significant. In the animals that died early, the control and high dose female mice show a positive correlation between the two tumors, though the relationship is only statistically significant for the control animals. The Mantel-Hentzel test (Rosner, 1995) can be used to control for animal survival. It can be used to combine 2x2 tables 3b and 3c to test statistical significance of the relationship between occurrence of liver adenomas and carcinomas. Combining scheduled sacrificed and early died animals by this test reveals similar anticorrelation between adenomas and carcinomas (Table 3(d)).

We next looked at the correlation between liver adenomas and carcinomas in animals from experiments in which chemical administration was associated with statistically significant increases in either liver adenomas or carcinomas. Tables 4 (a)-(d) report the correlations in animals from high dose exposure groups of chemicals linked to an increase in adenoma rates. In every case there is a very strong and statistically significant deficit in carcinomas when adenomas are present. When we focus on animals from high dose groups in experiements yielding increases in liver carcinomas (Tables 5 (a)-(d)) we also find strong anticorrelations.  In both tables 3 and 4 the animals surviving to scheduled sacrifice show stronger anticorelations than the animals that died early.  However, the rates of adenomas and carcinomas P₇ and P₆₄ are similar in both groups. Combining scheduled sacrificed and early died animals by the Mantel-Hantzel test reveals statistically-significant anticorrelations between liver adenomas and carcinomas (Tables 4(d) and 5(d)).
 
 

3.2 Analysis of the relationship of animal body weight to the rate of liver adenomas or carcinomas based on CBDS database

There is a long-recognized relationship between animal body weight and likelihood of liver adenoma or carcinoma in untreated rodents (Turturro, 1996; Linkov et al., 1998b). We examined the relationship of animal body weight to the rate of liver adenomas or carcinomas to see if these two tumor types had the same, or different correlations with weight. We have shown previously that the control animals in the CBDS database have less temporal variability in body weight and a much weaker relationship between body weight and combined liver tumor rate than those from more recent bioassays of TDMS database (Linkov et al., 1998b). That emphasizes the importance of examining both the CBDS database and the TDMS database.

The CBDS database has information on the average weight of each group (contained in a single cage) during the experiment but lacks of information on individual animal body weight, except at the end of life. We first plotted end of life liver adenoma rates against average group weight at one year for all control male (Figure 1) and female (Figure 2) mice. Both plots show a positive relationship between group body weight at one year and adenoma rate at the end of life, although there is considerable variability in tumor rate for any particular weight. Similar plots for liver carcinomas (male mice Figure 3, females Figure 4) reveal a positive, but noisy, relationship for males but no relationship for females.

We next plotted the end of life individual animal body weights them versus tumor rates (Figures 5 and 6). Again, there is a strong and positive relationship between an animal’s end of life body weight and the probability of a liver adenoma. For carcinomas, the picture is different. In male mice, there is a strong negative correlation between end of life body weight and likelihood of carcinoma (Figure 7), while female mice (Figure 8) seem to have a positive relationship.

3.3 Lifetime mean body weight for control male and female mice based on TDMS database

As noted in the preceding paragraph, the CBDS database lacks information on individual animal body weight during the experiment. The TDMS database follows individual animal weights for the full duration of the experiment. We plotted mean body weight over the lifetime of all control animals, those that will develop liver adenomas, those that will develop liver carcinomas, and those that live a tumor free life, for male (Figure 10) and female (Figure 11) mice.

Figure 9 shows that male mice that develop liver tumors are larger, on average, than the average of all animals and those that do not develop any tumors, beginning quite early in life. The difference is noticeable at 20 weeks, well before tumors arise. Animals that develop carcinomas show dramatic decrease in mean body weight at the end of life, presumably due to the severe physiologic effects of the tumor.

The results for female mice (Figure 10) are similar although the weight difference between animals that will develop liver tumors and those that do not develop any tumors is smaller. This seems to be due to the fact that female mice that do not develop tumors are heavier than the average female, in contrast to male mice.

In figure 10 the line for the category "all animals" lies always below the lines for both the group of animals with liver tumors and the group of animals without any tumors. It therefore follows that a line for the group of animals not on this graph, namely, those that develop tumors other than liver tumors, must be lie appreciably below the line for the average of all animals showing that their weights must be smaller.

3.4 Survival of animals with or without tumors at the end of life

We also plotted the survival curves for all control male (Figure 11) and female (Figure 12) mice in the TDMS database, grouped by the same categories as in the plots of lifetime mean body weight. In male mice (Figure 11), the survival curves of animals that will develop liver adenomas and those that will develop liver carcinomas are above the curve of "all animal" category. Survival curve of animals that live a tumor free live is below those of the other three groups almost throughout the whole time period. One  possible explanation is that an appreciable  fraction of animals with no tumors are those that die early before tumors develop. The percentage of survivors in the group with liver carcinoma drops rapidly after 70 weeks. This drop shows the lethal effect of liver carcinoma on animals.

The plot of survival curves for female mice (Figure 13) shows similar results although the difference in percentage of survivors between animals without any tumor and animals with liver adenomas or carcinomas is smaller than that observed in male mice.

Discussion and Conclusions

This study began with a simple aim; to examine whether our practice of distinguishing between liver adenomas and carcinomas in rodent bioassay analysis is sound. We have been criticized for the practice (J. Haseman, personal communication) on the grounds that it is contrary to standard NTP analysis protocol (McConnell et al., 1986). One can imagine two reasons for not distinguishing the tumor types. First, it may be that pathologists cannot reliably distinguish between them and therefore tumors may be assigned inconsistently to the categories of adenoma and carcinoma between bioassays. We cannot address this concern but note that there is a large literature on the differences and similarities between these tumor types that relies upon distinctive pathological diagnosis. The second reason for combining the tumor types would be a belief that they behave the same biologically, having similar relationships with various attributes of a bioassay. This is the focus of our analysis.

In this study we find two key biologically relevant relationships. First, in a given animal, the likelihood of developing an hepatocellular adenoma or carcinoma is not independent. Second, the two tumor types have different relationships with body weight in control animals.

The anticorrelation between liver adenomas and carcinomas in mice is quite striking, although not totally consistent. In control animals there is a strong negative correlation in male mice, but a positive correlation in females. The reason for this is not clear although it is notable that control female mice have much lower rates of both adenomas and carcinomas. In all chemical-exposed groups, the negative correlation is apparent and strong.

One possible explanation for an anticorrelation (negative correlation) of liver adenomas and carcinomas is the lethality of one tumor type. If animals that develop the more lethal tumor die early, they simply may not live long enough to develop the other tumor. We divide the animals in our database into two groups, those surviving until the scheduled end of the study (scheduled sacrifice) and those that die early (excluding scheduled sacrifice). If survival differences account for the negative correlation, we would expect to see a strong negative correlation between adenomas and carcinomas (i.e. odds ratios less than one) in the group of animals that died early while surviving animals should show no correlation (or even positive to offset the effects of the early death group). In fact, we see no such result and even find smaller odds ratios (and conditional probabilities) in the scheduled sacrifice animals. The exception again is control female mice in which scheduled sacrifice animals begin to look like all other groups with an odds ratio less than one, though not statistically significant. It is interesting to note that the rates of each tumor type do not differ significantly between animals that survive until scheduled sacrifice and those that do not. We cannot exclude differences in pathologic diagnosis in the two groups as a possible explanation. One can imagine that the occasional animal dying early might have a more thorough necropsy than one of the large number that are processed following scheduled sacrifice. If a more complete necropsy is more likely to detect all tumors in the liver, animals dying early might appear to have a different correlation than those examined after scheduled sacrifice.

When we focus on animals from the high dose groups of those experiments in which chemical treatment led to statistically significant increases in either liver adenomas or carcinomas, the anticorrelation between the tumor types persists. Some of the tumors in these high dose groups are caused by the chemical but others will be caused by the natural backround processes that are also present in the control animals. Therefore the anticorrelation observed is a combination of the observed anticorrelation in the backround and a correlation (either positive or negative) in the chemiaccly induced tumors. If there were no correlation or anticorrelation in the chemically induced tumors , groups of animals with chemically-induced increases of one or the other should have odds ratios tending back toward one. We see, in fact, a stronger relationship in animals with induced tumors, especially in the animals living to scheduled sacrifice. This suggests that for chemically induced tumors there is an even larger anticorrelation and may be further support for the notion (Butler, 1996) that adenomas and carcinomas are different steps on a continuum of liver neoplasm.

It is clear that the body weight of a mouse is related to the likelihood of developing liver tumors (Turturro et al., 1996; Linkov et al., 1998b). We have demonstrated that the relationship is different, however, for hepatocellular adenomas and carcinomas. Adenoma rates show a strong positive association with group average body weight at one year. The carcinoma picture is more complicated. The relationship is positive for control male mice but essentially nonexistent for females. Again, in both cases the tumor rates are much lower in females. End of life weights also show differences between liver adenoma and carcinoma, with the adenoma results mirroring the one year body weight relationship. The carcinoma relationship is again mixed with heavier males having a lower rate of carcinoma. The simplest explanation is that animals with carcinomas are too sick to be at the high end of the weight scale. The relationship in females is less stark but generally similar to the males except for the heaviest animals.

When we evaluate the body weight over a lifetime a fascinating fact arises, animals that will go on to develop tumors are heavier than those that will not from a young age, presumably well before a tumor could alter physiology or weight. The data also supports the notion that animals with carcinoma are quite sick and lose weight at the end of life.

Further support for survival differences not influencing the negative correlation come from figures 11 and 12. Apparently, animals with tumors actually live longer than the average animal. One possible explanation is that some fraction of the animals with no tumors are those that die early, before any tumor can become apparent. Another possibility is that other tumors, included in the "all animals" category, are more lethal than liver tumors.

This exercise has shown the value of separately considering liver adenomas and carcinomas in analysis. Distinct patterns can be ascertained and biological relationships explored. We hope that others will be encouraged to preserve the pathological distinctiveness of the two tumor types, in spite of possible difficulty in diagnosis, when analyzing rodent bioassays.

Acknowledgements

The authors are grateful to the members of the National Toxicology Program for making the CBDS data available (in 1986) and the TBDS data available (in 1999).   This work was begun with a grant from the Pardee Foundation, and completed using gifts to Harvard University from Zeneca Corp and Pfizer Inc. and private individuals.

References

Butler, W.H. (1996) "A Review of the Hepatitic Tumors Related to Mixed-Function Oxidase Induction in the Mouse"  Toxicological Pathology 24:484-492

Byrd, D., E.A.C. Crouch, and R. Wilson (1990) "Do Mouse Liver Tumors Predict Rat Tumors: A Study of Concordance Between Tumors Induced in Different Sites in Rats and Mice." Mouse Liver Carcinogenesis: Mechanism and Species Comparisons. p. 19-41

Gart, J.J., et al. (1986). Statistical Methods in Cancer Research. International Agency for Research on Cancer, Lyon.

Gold, L.S., T.H. Slone, N.B. Manley, and L. Bernstein (1991) "Target Organs in Chronic Bioassays of 533 Chemical Carcinogens." Environmental Health Perspectives 93:233-246

Hart, R.W., Keenan, K., Turturro, A., Abdo, K.M., Leakey, J., and Lyn-Cook, B. (1995a). Caloric restriction and toxicity. Fundam. Appl. Toxicol. 25, 184-195.

Hart, R.W., Turturro, A., Leakey, J., and Allaben, W.T. (1995b) Diet and test animals. Science 270, 1419-1421.

Haseman, J.K., Young, E, Eustis, S.L., and Hailey J.R. (1997) Body weight-tumor incidence correlations in long-term rodent carcinogenicity studies. Toxicol Pathol 25 :256-63

IRIS (1999) EPA’s Integrated Risk Information System http://www.epa.gov/ngispgm3/iris/index.html

Keenan, K.P., Laroque, P., Ballam, G.C., Soper, K.A., Dixit, R., Mattson, B.A., Adams, S.P., and Coleman, J.B. (1996). The effect of diet, ad libitum overfeeding, and moderate dietary restriction on the rodent bioassay: the uncontrolled variable in safety assessment. Toxicol. Pathol. 24, 757-768.

Linkov, I., Wilson, R., and Gray, G.M. (1998a). Anticarcinogenic responses in rodent cancer bioassays are not explained by random effects. Toxicol. Sci. 43, 1-9.

Linkov, I., Wilson, R., and Gray, G.M. (1998b). Weight and survival depression in rodent bioassays with and without tumor decreases. Toxicol. Sci. 43, 10-19.

Linkov, I., Wilson, R., and Gray, G.M (1999) "Errata in above two papers Toxicol. Sci

Linkov, I., Polkanov M. Wilson, R., and Gray, G.M (1999) Correlation among tumor types in Cancer Bioassays: Liver adenomas, Liver Carcinomas, Leukemias and Lymphomas (in process of publcation)

McConnell, E.E., Solleveld, H.A., Swenberg, J.A., and Boorman, G.A. (1986) Guidelines for Combining Neoplasms for Evaluation of Rodent Carcinogenesis Studies. JNCI 76:283-289

Rosner, B. (1995). Fundamentals of Biostatistics. Duxbury Prese provided to us by Dr Angelo Tut of individ

Turturro, A., Duffy, P., Hart, R.W., and Allaben, W.T. (1996). Rationale for the use of dietary control in toxicity studies - B6C3F1 mouse. Toxicol. Pathol. 24, 769-775.

Ward, J.M., Shibata, M., Devor, D.E. (1996). Emerging Issues in Mouse Liver Carcinogenesis. Toxicol. Path 24, 129-137.

Young S.S., and Gries C.L. (1984) Exploration of the negative correlation between proliferative hepatocellular lesions and lymphoma in rats and mice--establishment and implications. Fundam Appl Toxicol 4:632-40
 
 
 
 
 
 
 
 

Table 3. Correlations Between Liver Adenomas (Classification 7) and Carcinomas (Classification 64) for Mice

3 (a) all animals
 
 

Animal Group	Sex	+/+	+/-	-/+	-/-	OR	c2	+/+ Expect	P7	P64	P64|7	P64|7 / P64
All Controls	M	116	1242	2461	11395	0.43	74.72	230	0.09	0.17	0.09	0.50
All Low Dose	M	165	1467	3388	11535	0.38	138.41	350	0.10	0.21	0.10	0.47
All High Dose	M	191	1386	2952	9835	0.46	98.91	345	0.11	0.22	0.12	0.55
All Controls	F	24	487	467	15224	1.61	4.98	15	0.03	0.03	0.05	1.55
All Low Dose	F	36	816	1380	14662	0.47	20.18	71	0.05	0.08	0.04	0.50
All High Dose	F	64	875	1474	12076	0.60	15.27	100	0.06	0.11	0.07	0.64

 

3 (b) scheduled sacrifice only
 
 

Animal Group	Sex	+/+	+/-	-/+	-/-	OR	c2	+/+ Expect	P7	P64	P64|7	P64|7 / P64
All Controls	M	32	611	1377	6365	0.24	69.68	108	0.08	0.17	0.05	0.30
All Low Dose	M	59	762	1997	6276	0.24	122.67	186	0.09	0.23	0.07	0.32
All High Dose	M	37	725	1522	4599	0.15	154.86	173	0.11	0.23	0.05	0.21
All Controls	F	6	232	307	9229	0.78	0.37	8	0.02	0.03	0.03	0.79
All Low Dose	F	13	473	885	8200	0.26	27.09	46	0.05	0.09	0.03	0.29
All High Dose	F	13	490	789	5956	0.20	39.50	56	0.07	0.11	0.03	0.23

 
 
 

3 (c) excluding scheduled sacrifice
 
 

Animal Group	Sex	+/+	+/-	-/+	-/-	OR	c2	+/+ Expect	P7	P64	P64|7	P64|7 / P64
All Controls	M	84	631	1084	5030	0.62	16.15	122	0.10	0.17	0.12	0.69
All Low Dose	M	106	705	1391	5259	0.57	27.75	163	0.11	0.20	0.13	0.65
All High Dose	M	154	661	1430	5236	0.85	2.84	173	0.11	0.21	0.19	0.89
All Controls	F	18	255	160	5995	2.65	15.49	8	0.04	0.03	0.07	2.38
All Low Dose	F	23	343	495	6462	0.88	0.37	26	0.05	0.07	0.06	0.89
All High Dose	F	51	385	685	6120	1.18	1.19	44	0.06	0.10	0.12	1.15

 

3 (d) combining scheduled sacrificed and died early animal groups by Mantel-Haenzel test
 
 

Animal Group	Sex	Observed	Expected	Variance	c2	Odds ratio	p-Value
All Controls	M	116	230	174	74.57	0.43	5.8*10-18
All Low Dose	M	165	348	247	135.53	0.39	2.5*10-31
All High Dose	M	191	345	240	98.39	0.46	3.6*10-23
All Controls	F	24	15	14	4.86	1.65	0.03
All Low Dose	F	36	71	62	19.72	0.47	9.0*10-6
All High Dose	F	64	100	83	15.07	0.60	1.0*10-4

 
 
 

Table 4. High Dose Groups from Bioassays with Statistically Significant Increase in Adenomas (classification 7). All Animals or Only Those Surviving to Scheduled Sacrifice (SS)

4 (a) all animals
 
 

Animal Group	Sex	+/+	+/-	-/+	-/-	OR	c2	+/+ Expect	P7	P64	P64|7	P64|7 / P64
High Dose	M	90	532	778	1729	0.38	68.21	173	0.20	0.28	0.14	0.52
High Dose	F	42	436	573	2165	0.36	38.79	91	0.15	0.19	0.09	0.46

 

4 (b) scheduled sacrifice only
 
 

Animal Group	Sex	+/+	+/-	-/+	-/-	OR	c2	+/+ Expect	P7	P64	P64|7	P64|7 / P64
High Dose	M	16	293	349	737	0.12	90.49	81	0.22	0.26	0.05	0.20
High Dose	F	7	242	296	948	0.09	56.47	51	0.17	0.20	0.03	0.14

 

4 (c) excluding scheduled sacrifice
 
 

Animal Group	Sex	+/+	+/-	-/+	-/-	OR	c2	+/+ Expect	P7	P64	P64|7	P64|7 / P64
High Dose	M	74	239	429	992	0.72	5.34	91	0.18	0.29	0.24	0.82
High Dose	F	35	194	277	1217	0.79	1.42	41	0.13	0.18	0.15	0.84

 

4 (d) combining scheduled sacrificed and died early animal groups by Mantel-Haenzel test
 
 

Animal Group	Sex	Observed	Expected	Variance	c2	Odds ratio	p-Value
High Dose	M	90	172	99