216 Diana Anderson individual values within groups of exposed people is wide and overlaps the ranges of adjacent or control groups. In retrospective studies where exposure has occurred for up to 20 or 30 years, it is often difficult to know exact exposure levels and exposure can often be only estimated roughly from occupation. We are investigating whether such a study is useful in trying to determine safe exposure levels. If the exposed population is not significantly different in terms of chromo- some damage from the control population, then it might be assumed that safe exposure levels have been achieved, because a negative result suggests that the chemical concerned is not a mutagen. However, there are limitations to this approach, since the exposure level may be too low to produce the chromosome damaging effect but may still cause sister chromatid exchanges or gene mutations which are not detected by conventional chromo- somal analysis. Controls should be taken from both on and off site where possible, and should be age and sex matched. Cells are generally cultured for 48 and 72 h and we have found no significant differences in data at these two times after vinyl chloride exposure (91), although after irradiation this is not the case (92) and 48 h cultures are desirable. At least 100 cells per individual should be analysed from slides coded to avoid observer bias. Whilst there is a correlation bteween carcinogenesis and mutagenesis (earlier references), a review paper by Hamden (93) puts the correlation for clastogenic and carcinogenic effects into perspective, as does the book edited by German (94). There appears to be a fairly good but non-quantitative correlation. Other techniques such as sister chromatid exchange may be useful on exposed workers, but effects are much more short-lived (95, 96). and may to some extent have disappeared before culturing is possible. Once it is established that a chemical is clastogenic, regular population monitoring of the work force may be initiated, in which the workers are monitored both pre-exposure and during employment The results of the monitoring will be useful for checking plant hygiene, and an increase in absormal cells in an individual could be used as an indication that the worker should not continue to be exposed to that chemical. If an individual is found to have a chromosomal abnormality linked to a certain disease he can be advised through appropriate medical channels of the risk of inheritance of the disease in any children he may have. If all workers in chemical plants were monitored as part of a routine medical surveil- lance service, then many of the difficulties involved in initiating prospective studies would disappear. Computerised microscopic techniques are available which can reduce the time spent by a technician by about 40•o. At present, however, metaphase spreads are merely located and the amount of damage still has to be assessed visually. Initial cost of purchas- ing a computer microscope is very high. Other techniques for direct application to man are available, such as the use of urine or blood plasma from exposed workers in combination with a microbial assay (97), electrophoretic monitoring of enzymatic markers in man (98), detection of variants in haemoglobulin molecules (99-100), investigations of sperm morphology (101) and an increase in the presence of YY bodies (102). INTERPRETATION OF DATA Assuming that experiments are reproducible, in well-conducted experiments using

The current state of mutagenicity testing 217 adequate testing protocols, interpretation of data may still be difficult especially with results within the control range or on the border-line of statistical significance. Inter- pretations may differ depending on whether results from treatment groups are compared with historical (accumulated) or concurrent controls, the type of statistical analysis or whether results are compared at equitoxic doses. If the experiments are repeated, similar equivocal results may be obtained and if not there is the problem of whether the first or repeat experiment is correct. The difficulty is in proving and accepting negative data. By comparison the handling of positive data is much more clear-cut. Systems giving negative results are often considered insensitive, e.g. the dominant lethal test. However, such a test might truly reflect the situation in the germ cells and the testes may really be incapable of metabolising many compounds to their active inter- mediates. Microbial systems need large concentrations of compounds to detect a positive result by comparison with concentrations expected to be present in whole mammals, and thus could equally be considered insensitive. ASSESSMENT OF GENETIC EFFECT Making a quantitative risk evaluation for man from data obtained in testing procedures is an even more difficult task. First it must be determined if the data are biologically and/ or statistically significant. Chemicals with unknown mutagenic potential are probably more difficult to assess than data for radiation or known mutagens (many of which are 'radio-mimetic' agents) unless there are clear-cut dose responses, since chemicals behave differently with different cells, organs and organisms. One of the recommended ways of assessing risk is by comparing the values obtained from chemical data with corresponding radiation data. Crow (10) recommended that the population genetic effects of clinical mutagens be assessed, taking radiation as an equi- valent and equating the population dose of chemical mutagens to the radiation dose admissible for that population. Bridges (43, 44) in his papers describing the three-tier system also recommends the principle of a radiation-equivalent dose. With this system only those substances which have passed through the first two tiers and are of great social, medical and economic importance are subjected to a quantitative assessment. Chemicals that have shown a positive effect in the first two tiers should, if not widely used, be prohibited or used with a non-mutagenic derivative substituted for the mutagenic radical. Those that are positive in the third tier should be expressed as the equivalent of a radiation dose producing the same effect, and this makes it possible to standardise the level of chemical mutagens to the limits of the level of ionising radiation. However, with extrapolation back to very small doses it would be diffcult to decide which is the line of best fit, and with some chemicals there may be shoulders of varying size on dose response curves due to permeability problems with a chemical or other unknown factors. Yet another 'radiation-equivalent concept', the ABCW model (101), is based on the hypothesis that the radiation induced mutation rate per radiation unit and gene locus in a variety of organisms from microbes to mammals is proportional to their genome size. This model has been extended to chemical compounds (104). As a result it would seem possible to estimate risks for man on the basis of an upward extrapolation. However, this model has been criticised by Sankaranarayanan (105). The problem with these radiation concepts is that chemicals may behave differently from radiation induced free radicals and often have species type specificity. This problem has been highlighted by Auerbach (106) and Sobels (107).