214 Diana Anderson levels. Absolute increases in induced colony numbers are observed over some of the dose range above control values- which is not always the case for human cell systems, where mutants may be very sensitive to the inducing agent and are being killed more rapidly than non-mutant cells (79, 70). Bone marrow in Mammals This is a useful system in that no auxiliary metabolising system is required, animals can be dosed with compounds directly and bone marrow cells do not need stimulation to divide since they have a high proliferative cell activity (80-83). Rats are suitable species to use but others will do. With rodents, groups of 5-8 animals, 8-10 weeks of age, should be used per concentration of the test substance and positive control substance. A larger number of animals should be used in the negative control group to provide a better data base for comparison. A minimum of 50 cells, preferably 100, from each animal should be analysed from coded slides to avoid observer bias when scoring chromosome damage. The higher the background frequency, the higher the number of cells needed to be examined in order to detect a statistically significant effect (Buffier, personal com- munication). A maximum tolerated dose of compound should be used, together with a more realistic dose related to human exposure and an intermediate value. In a preliminary screen only the maximum tolerated dose need be used - or even a micronucleus test may be used (84, 85). Animals can be sacrificed at various time intervals after exposure. From our own studies and those of others (86, 81, 60) 6 h seems to be suitable time after multiple chemi- cal exposures and 24 h after single exposures. However, these times may not be suitable for all chemicals. The most relevant route of administration of the compound should be used. If man is exposed to the chemical by inhalation, then an inhalation route should be used. If he is likely to ingest a pesticide sprayed on crops, then an oral route is recommended either by gavage or in the diet. Rarely does man receive intraperitoneal or intravenous injections of compound except in the case of some medicines. Positive control animals should be housed under identical conditions to those dosed with the test compound of animals even if identical routes of administration of test com- pound and positive control substances cannot be achieved. Cells should be scored for all type of chromosome aberrations including chromosome gaps. We have shown with several positive control mutagens such as ethyl methane- sulphonate, mitomycin C, benzene and vinyl chloride that gaps are at least as sensitive an indicator of damage as other types of damage: they may indicate a toxic event as opposed to or in addition to a genetic event (Anderson and Richardson unpublished). Dominant Lethal Mutation Assay This is one of the whole mammal tests on germ cells (87). The tests should be carried out on random-bred rodents of 8-10 weeks of age. The comments which apply to dosing regimes in bone marrow cells also apply to dominant lethal tests, i.e. a maximum tolerated dose should be used together with a dose related to human exposure levels and an inter- mediate value. Again, the most relevant route of administration should be used, and animals treated with the positive control substance sould be housed under identical conditions to those of animals exposed to the test compound.

The current state of mutagenicity testing 215 Each test group should consist of at least 10-20 males if males are exposed. (Domi- nant lethal effects in females are difficult to separate from systemic effects.) A pre-treat- ment fertility test should be undertaken to check fertility of the animals used and to determine a background dominant lethal frequency. Each male is bred with 2-4 females once a week, for 8 weeks in the case of mice or 10 weeks in rats, to sample all stages of the spermatogenic cycle. Negative control animal numbers should be larger, again to provide a better data base for comparison. (The size of the negative control group can be determined by the square root of the number of treatment groups, i.e. if there are 9 treatment groups each of 10 animals, then there should be 30 in the control group (88).) The statistics used in the dominant lethal test are complicated various statistical methods can be used, each with its own bias. The greater the number of statistical methods used, the greater the chance of producing a false positive result (Anderson, unpublished). Different evaluation methods are used for dead implant data. The US Food and Drug Administration has analysed all data on dead implants in two ways (both on dead implants/total implants/female basis and on dead implants/female) and infer that equally significant data are provided by both (89). If late deaths are eliminated from the analysis of the above parameters a more sensitive index of dominant lethality is obtained. Preimplantation losses can be indicated by comparing values of total implants in females mated with treated males and those mated with control males, as suggested by Epstein (90) instead of counting corpora lutea. Decreases in total implants which re- present increases in preimplantation losses without a corresponding increase in early deaths may not represent a mutagenic event. Other than genetic factors can explain a preimplantation egg loss. Fertility effects can also be determined in the dominant lethal test if fertile animals only are used for the study. The dominant lethal test is best con- ducted with at least three concentrations of test compound in order to try to obtain a dose-response relationship from which extrapolation of risk might be made. Clear-cut genetic effects are best claimed when a dose dependent increase in post-implantational foetal deaths is evident. Other methods such as the heritable translocation test detect transmitted damage and are useful is this respect. However, such a test is extremely expensive, as is the specific locus gene mutation test, requiring very large numbers of animals, extensive housing and maintenance. A negative result in any one of the systems so far discussed may mean that the wrong species of animals or microsomes has been used, or a result may be negative for any one of the reasons discussed earlier, or it may be a true negative. Human Peripheral Lymphocytes from Exposed Workers This sort of study is usually carried out retrospectively after a compound has been identified as a hazard, witness the many publications on workers exposed to vinyl chloride (91). Before such a study can be undertaken. many ethical problems have to be taken into consideration, e.g. if the exposed work-force need to be told the results of the study and what the results mean in the light of current knowledge. Such a study and decisions relating to it involve negotiations with medical officers, workers, unions, etc. Negotiation may be more difficult on a prospective basis where compounds present an unknown hazard. If a positive response is obtained, while it is possible to show a cor- relation between exposure and level of abnormal cells on a groups basis, the range of