Carcinogenicity Evaluation: Comparison of Tumor Data from Dual Control Groups in the CD–1 Mouse

Abstract

Current regulatory thinking allows for the use of single control groups for rodent carcinogenicity testing although there has been a trend until recently to use dual control groups. To date, virtually nothing has been published on whether a shift from dual to single control groups will affect the identification of tumorigenic risk potential in these studies. A recent evaluation of dual control carcinogenicity data in the rat (Baldrick, Toxicol Pathol 2005, 33: 283–291) showed that although no major differences in tumor incidences between the control groups were found, some interstudy variation occurred and in cases were a notable difference was seen, the use of 2 control groups, as well as robust, contemporary background data, allowed an easier interpretation of findings in drug-treated groups. In this paper, the results of 10 mouse carcinogenicity studies, performed between 1991 and 2004, with 2 control groups, are presented. As in the rat, interstudy variation was seen and in some cases, the use of dual control groups assisted in the tumor risk assessment. Thus, the continued use of 2 control groups can have a vital role in mouse carcinogenicity studies. The paper also presents an update on survival, on the range and extent of background spontaneous neoplasms and comments on genetic drift in this commonly used mouse strain.

Introduction

As part of the drug development process, rodent carcinogenicity studies are required by regulatory agencies for products that will have chronic clinical use or for which a particular class or mechanistic safety alert has been identified. Over the years, major region-specific (e.g., Europe, United States and Japan) differences in the design of these studies have been eliminated by the emergence of guidance documents arising from the International Conference on Harmonization (ICH), namely ICH S1A, ICH S1B, ICH S1C and ICH S1C(R). However, such guidance does not address whether one or 2 control groups need to be included as part of the study design.

Until recently, the regional situation gave a mixed picture: in Europe 2 control groups dosed with vehicle was indicated (Guidelines Volume 3B, 1998), United States guidance state one concurrent control group (FDA Redbook, 2000), which agrees with OECD requirements (OECD, 1993) and in Japan, a negative control group is mentioned (Japan, 1999). However, due to recent clarification in Europe that one control group of the same number for each sex dosed with the vehicle by the same route, is acceptable (CPMP/SWP/2879/00), it seems that “globally,” regulators will now accept the single control group model. It remains to be seen whether drug companies that have routinely used double control groups, whether for regulatory compliance, scientific or traditional reasons, continue with this trend.

Although carcinogenicity studies are performed to identify tumoriogenic potential in animals as part of the assessment of risk of use in humans, little published data are available on how the actual use of 2 control groups has assisted this process. A recent review of 13 carcinogenicity studies in the rat with dual control groups showed that on occasion, the presence of both groups assisted in the interpretation of tumor findings (Baldrick, 2005). Additional controls are reported as giving a better assessment of how rare each tumor type is and allows for a more accurate comparison to historical control data (Fairweather et al., 1998). It is also said that the results from 2 identical controls can be used as a mechanism for identifying the extent of control variability and to help evaluate the biological significance of increases in tumor incidence in the treated groups (CDER, 2001).

This paper presents data from 10 in-house carcinogenicity studies using CD–1 mice with 2 identical control groups. All the studies involved oral administration and were performed in the period 1991–2004. As well as examining potential tumor incidence differences between the dual controls, the paper also provides an update on the range and extent of background neoplasms in this commonly used mouse strain and will comment on genetic drift.

Materials and Methods

Studies for evaluation from our in-house database were limited to carcinogenicity investigations in gang-housed CD–1 (Crl:CD–1(ICR)BR) mice with 2 identical control groups using oral administration (gavage tube or dietary) in the period 1991–2004. A total of 10 studies were identified giving a total of 564 males and 564 females for each control group. Group sizes ranged from 51–60 animals/sex. Treatment comprised gavage dosing at 10 mL/kg of vehicle material or dietary intake of standard chow. Mice were treated for an approximately 2-year period (up to 105 weeks).

Mice of approximately 6–8 weeks of age at the start of dosing were obtained from Charles River UK and were generally housed 3 to a cage (one study had males housed singly and one study had both sexes housed singly). They were fed SQC Rat and Mouse Maintenance Diet No. 1 (Special Diets Service Ltd., United Kingdom) ad libitium and had free access to water. Animals were maintained on a 12h light/dark cycle at 19–25°C and 40–70% relative humidity. At the end of the treatment period, complete necropsies were conducted on the control animals and tissues were fixed and embedded in paraffin wax, sectioned (nominally 5 μm) and stained with hematoxylin and eosin. Tissues routinely examined were adrenals, brain, cecum, colon, duodenum, eyes, femur with bone marrow and articular surface, gall bladder, gross lesions, Harderian glands, heart, ileum, jejunum, kidneys, lacrimal glands, liver, lungs with mainstem bronchi, mammary gland (female), mandibular and mesenteric lymph nodes, muscle, esophagus, optic nerve, ovaries, pancreas, pituitary, prostate, rectum, salivary glands, sciatic nerves, seminal vesicles, skin, spinal cord (at 3 levels), spleen, sternum with bone marrow, stomach, testes + epididymides, thymus, thyroids + parathyroids, tissue masses, tongue, trachea, urinary bladder, uterus and vagina.

Survival after 2 years of treatment was measured. Mice were examined for non–neoplastic and neoplastic findings. Tumor identification was based on in–house nomenclature which followed that commonly used elsewhere, e.g., International Agency for Research on Cancer (IARC) and Society of Toxicologic Pathologists (STP). To allow for in–house consistency, “older” tumor synonyms have been used for some neoplasms. Statistical evaluation was performed for differences between the 2 control groups (designated control I and control II). Survival and tumor incidence was compared using a 2-sided chi-squared test. If the incidences were shown to be not significantly different, then the 2 groups were combined for comparison with drug-treated groups in each of the presented carcinogenicity studies. All studies were performed to Good Laboratory Practice (GLP) and animal care conformed to the United Kingdom Animals (Scientific Procedures) Act of 1986.

Results

Control group survival at study end is given in Table 1. Overall, mean survival was similar between the 2 groups with 30–72% and 27–73% survival seen for males and 25–78% and 33–65% survival seen for females in control groups I and II, respectively. Generally, only small differences between the 2 control groups were seen for individual studies with no statistically significant differences in survival. When survival was broken down to 7-year periods of 1991–1997 and 1998–2004 (the most appropriate split based on the dates when the studies were performed), there was no obvious time effect on survival values. Across all studies, the main causes of death were consistent with those generally associated with aging laboratory mice of this strain and comprised amyloidosis (both sexes) and urogenital tract lesions (predominantly in males) as well as tumors of the hemolymphoreticular system, lung and skin/subcutis (both sexes) and liver (predominantly males).

Common nonneoplastic lesions in decedents and mice surviving to termination are listed in Table 2. There were no obvious group differences seen across the studies. The findings were consistent with the spectrum associated with aging mice of this strain and included amyloid deposition, arteritis and inflammatory cell foci across various tissues and nephropathic findings.

A summary of tumor incidences is given in Table 3 with details of major organ or common and less common tumors shown in Tables 4 and 5, respectively. Notable interstudy differences in tumor incidences between dual control groups are presented in Table 6. Table 3 shows that the total number of neoplasms was similar to each other for male and female control groups; benign tumors occurred with approximately twice the frequency of malignant tumors in males but with a fairly similar incidence to malignant tumors among females. Across all the groups, the malignant tumor incidence was 30–34% (males) and 53–56% (females). When tumor incidences were broken down to 7-year periods of 1991–1997 and 1998–2004, results were generally similar although the incidence for females was slightly increased (through malignant tumor burden) in the most recent 7-year period compared to the earlier time period.

As seen from Table 4, the most common tumors were those associated with the digestive system (comprising hepatocellular adenomas and/or carcinomas in males), the respiratory system (comprising bronchiolo-alveolar adenomas or carcinomas with a higher incidence of adenomas in males compared to females) and the hematopoietic/lymphoreticular system (comprising lymphomas [lymphocytic or pleomorphic] with a higher incidence in females compared to males). Other more frequent tumors were those of the reproductive system (comprising uterine stromal polyps, leiomyomas or histiocystic sarcomas) and the integumentary system (comprising mammary gland adenocarcinomas in females). Remaining tumors (when assessed from a full complement of mice) were present at <5%.

Overall, “notable” differences in the number of individual tumors were only seen between the control groups (control I vs. control II) on a few occasions among the combined data for the 10 studies examined, most notably as follows:

• adrenal pheochromocytomas in females (1 vs. 4);

• pancreatic islet cell adenomas in males (5 vs. 0) and females (4 vs. 1);

• skin fibrous histiocytomas in females (7 vs. 2);

• uterine leiomyomas (19 vs. 29);

• uterine histiocytic sarcomas (25 vs. 15);

• hepatocellular adenomas in females (5 vs. 9);

• hepatic histiocytic sarcomas in males (0 vs. 3);

• lymphoma lymphocytic in males (10 vs. 18);

• granulocytic leukemias in males (6 vs. 19) and females (12 vs. 6);

• hematopoietic/lymphoreticular histiocytic sarcomas in males (2 vs. 6) and females (9 vs. 3);

• Harderian gland adenomas in males (13 vs. 20).

As seen from Table 5, there was no specific pattern of incidence for less common tumors which, in the most extreme case, were only seen on one occasion among all the animals examined. It should be noted, however, that there were various occasions when a rare tumor was seen in 2 animals in one control group and not at all in another group. In agreement with the overall variation between the 2 control groups, all 10 studies individually showed cases of “notable” differences as can be seen from Table 6; these differences tended to be among the more commonly seen tumor types. As will be pointed out in the Discussion, it is the magnitude of the difference with relevance to findings in drug-treated groups and in relation to background data that are important.

Discussion

This paper shows that, in general, no major differences occur for any tumor incidence between dual control groups across the full range of mammalian body systems when assessed in 10 in-house mouse carcinogenicity studies over a period of 14 years. Interstudy variation occurred for some tumors but did not follow any particular pattern. A similar result was found in a recent evaluation of dual control carcinogenicity data in the rat (Baldrick, 2005). Unfortunately, due to the general lack of publications in this area it is difficult to assess whether other authors have considered such interstudy variation important.

An evaluation of the Food and Drug Administration (FDA) and National Toxicology Program/National Cancer Institute (NTP/NCI) database from 1979 to 1992 showed that most rodent carcinogenicity studies had one control group and only 18 cases of duplicate control groups were mentioned (Contrera et al., 1997). These latter studies appear to be a series of experiments performed with 18 color additives in the CD–1 mouse (Haseman et al., 1986). Review of the findings by these latter authors showed marked study-to-study variability for certain tumors with statistically significant (p < 0.05) paired control differences in 7 of the studies (see Table 8).

However, when these results were analyzed to see if these findings exceeded chance expectation, no evidence of extra-binomial within-study variability between the 2 concurrent control groups was found. Indeed, the total number of observed significant paired-control differences was virtually identical to what would be expected from usual binomial model assumptions. Examination of dual control data in CD–1 mice using an unidentified chemical by Selwyn (1989) showed 3 cases of statistically significant differences (p < 0.05) among tumors (see Table 8). Upon analysis, it was concluded that statistically significant differences between identically treated groups will occur with regular frequency but such data do not provide strong evidence of extrabinomial variation in tumor rates. Overall, these findings support the view that caution is needed when interpreting statistically significant findings.

It would appear that although some interstudy variation can be expected in carcinogenicity study control group tumor incidences, such variation does not cause an issue in assessment for potential carcinogenic risk in most cases. This situation is supported by the fact that dual control tumor data are usually added together to allow comparison with drug-treated groups. Where dual control data come into their own is when variation (as seen in the present study between concurrent control groups of for example, 3 vs. 11 systemic lymphomas for males in one study and 0 vs. 4 pituitary adenomas for females in another study) can assist interpretation of an increased incidence of the tumor in the high drug-treated group that would be “diluted” if the 2 control group data are summed.

As indicated earlier, it is not necessarily important if the difference in tumor incidence between dual controls is statistically significant in itself; what is important is the magnitude of the difference in conjunction with factors such as preneoplastic/hyperplastic changes, evidence of increasing tumor incidence with increasing dose and background data incidence. However, such an argument can work both ways. In the case of an incidence of 8/51 vs. 2/51 hepatocellular adenomas in males for the 2 control groups in one study, an incidence of 8/51 in the high drug-treated group could be easily discounted as non–treatment-related and that one control group would have supported such a conclusion as the incidence was within the background data range.

In the case of an incidence of 4/51 vs. 0/51 pancreatic islet cell adenomas in males in the concurrent controls in another study, it would be more difficult to explain an incidence of for example, 4 such tumors in the high-dose group in the absence of both control groups as background data may not have been useful (Table 4 shows that only 5/452 males in Control I and 0/421 males in Control II had this tumor). The use of dual control groups to support interpretation of rare tumors is also valid, although results from the presented 10 carcinogenicity studies indicate that such tumors were often seen in both control groups anyway.

Overall, in all these cases, scientific assessment would involve a careful evaluation of tumor incidence across all study groups (control and drug-treated) with reference to contemporary mean and overall range tumor values. Indeed, it has been pointed out elsewhere that 2 control groups within the same study can occasionally show notable differences, stressing the importance of knowing the range of incidences rather than just mean incidence (Gopinath, 1994). Results from the present investigation show that the continued use of dual control groups in carcinogenicity studies has a role in the assessment of tumorigenic risk.

Various authors have reported survival, cause of mortality and/or spontaneous neoplasm data for CD–1 mice (e.g., Maita et al., 1988; Chandra and Frith, 1992; Criver, 1995; Gopinath, 1994; Son, 2003; Son and Gopinath, 2004; Criver, 2005; Hooks et al, 2005). Maita et al. (1988) showed survival of 20–44% (males) and 27–43% (females) from 11 carcinogenicity studies in the late 1970s/early 1980s while Chandra and Frith (1992) reported values of 41% (males) and 47% (females) from a survey of 11 in-house studies largely conducted in the late 1980s; further data from 9 studies of 2-year duration again largely from the late 1980s showed survival ranges of 28–72% (males) and 27–65% (females) (Criver, 1995). Further authors report survival of 48% (males) and 44% (females) from a range of studies in the period 1995–2001 (Hooks et al, 2005) plus survival values of 18–66% (males) and 19–55% (females) from 14 studies of 2–year duration from 1990–2000 (Criver, 2005). These latter values compare well with the survival range of 27–73% (males) and 25–78% (females) or overall survival means of 43% (males) and 42% (females) seen in the present reported studies.

The major factors contributory to death in the present study were amyloidosis and urogenital tract lesions as well as tumors of the hemolymphoreticular system, lung and skin/subcutis and liver. These factors compare favorably with those reported elsewhere in CD–1 mice (e.g., Maita et al., 1988; Chandra and Frith, 1992; Son, 2003; Son and Gopinath, 2004; Hooks et al., 2005).

The present investigation shows that benign tumors were just over twice as common as malignant tumors in males (68 vs. 32%) while a fairly similar incidence occurred in females (46 vs. 54%); these values are similar to the 70 vs. 30% (males) and 55 vs. 45% (females) incidence, respectively reported elsewhere (Chandra and Frith, 1992). Surprisingly few publications cite actual frequency of neoplasms but as can be seen from Table 7, a fairly consistent pattern of common tumors has been reported for the CD–1 mouse with the most frequent being hepatocellular adenomas and/or carcinomas in males, bronchiolo-alveolar adenomas and carcinomas in both sexes and systemic lymphomas which occurred at a higher incidence in females. A similar finding has also been reported by Hooks et al. (2005).

Uterine tumors (notably stromal polyp and leiomyomas), mammary gland adenocarcinomas and pituitary adenomas were also at a frequency of 2–6% across the literature and the present investigation. Harderian gland adenomas were also seen often in control animals but calculation of an overall background incidence is made difficult as, on occasion, not all the mice available in the data sets were examined (e.g., only about 25% of available mice were actually examined for this tumor type in the present study). As pointed out elsewhere for Harderian gland neoplasm evaluation, some tissues are not routinely included in pathology protocols but are only examined when grossly abnormal (Hardisty, 1985). Values for less common tumors were also found to compare well with those given in the published literature. Interestingly, no specific tumors of the nervous or cardiovascular systems were detected which confirmed the rare nature of such neoplasms reported elsewhere (e.g., Maita et al., 1988; Chandra and Frith, 1992; Criver, 1995, 2005).

Overall, the general similarity of tumor incidence in the present study to those reported elsewhere for CD–1 mice is comforting bearing in mind probable differences in study design. Such differences include mouse strain/supplier, diet (composition and amount), caging (single vs. gang housed), environmental conditions and pathology criteria. The lack of any major shifts in tumors incidences across different investigations raises the question of what role genetic drift may have. Gopinath (1994) has reported that variation of incidence of tumor types within the same strain and sex and, within the same laboratory, occur due to the progression of time. In order to avoid such “genetic drift” when using historical data, it has been suggested that only studies performed during the previous 3–4 years (Haseman et al., 1984; Fairweather et al., 1998), 3–5 years (Greim et al., 2003 referring to the European Registry of Industrial Toxicology Animal database), 5 years (CPMP/SWP/2879/00) or 7 years (Haseman et al., 1998) prior to the new study, should be accessed.

From an analysis of certain tumor types over a period of 9 years (under conditions were the potential sources of variability were recognised and strictly controlled), Gopinath (1994) showed that although there were fluctuations, incidence rates remained within a small band and no clear evidence of a time-related drift in rates was found. Furthermore, a review of the historical control data for 20 carcinogenicity studies generated in 1990–2002 in CD–1 mice showed little variability with time (Son and Gopinath, 2004). An evaluation of tumor incidence using 7-year time points over the 14-year period in the present investigation has also shown no obvious difference in the number or pattern of tumors recorded. Standardizing tumor burden by incidence in the total number of animals assessed has not shown any strong evidence of drift although a slightly higher incidence of tumors for females was seen in the most recent 7–year period.

In conclusion, this paper shows that no major differences in tumor incidences occur when dual control groups are used for mouse carcinogenicity studies. However, occasional interstudy variation occurred for some tumors indicating that the use of 2 control groups has a role in the interpretation of findings in drug-treated groups. Such interstudy variation assists interpretation when one control group, but not the other, mirrors elevated tumor incidence in one or more of the drug-treated groups. Thus, dual control groups give greater reassurance in assessing for biological significance of any increased tumor incidence. Such use also allows the buildup of contemporary background data at a greater rate.