International reference study on the identification and scoring of human chromosome aberrations

Bull. Org. mond. Sante 11974, 50, 373-388 Bull. Wid Hlth Org. j International reference study on the identification and scoring of human chromosome aberrations Results of a WHO comparative study * JOHN D. ABBATT,1 K. C. BORA,2 M. R. QUASTEL,3 & L. P. LEFKOVITCH4 Chromosome aberrations in human peripheral blood are recognized parameters of cellular damage and are used as indicators of exposure to ionizing radiation and certain chemicals. However, significant interlaboratory variability exists in the results reportedfrom different laboratories. The primary objective of the present study was to examine problems associated with the identification and analysis of chromosome aberrations. Significant interlaboratory variability was found to exist in the analysis ofhuman chromosome spreads for induced interphase aberrations, apparently owing inpart to differences in the selection and rejection ofspreads for scoring and also in the recognition and classification of various types of aberrations. These differences are reflected in the dose-response relationships for aberra- tions as well asfor damaged spreads. For damaged spreads the scoring variability appears to be relatively small. It is inferred that factors additional to differences in scoring may play a significant role in the large variation in the reported dose-response relationships both for ionizing radiation andfor chemically induced aberrations. Chromosome aberrations are recognized para- meters of cellular damage caused by ionizing radia- tion. They are readily quantifiable and when classi- fied can be interpreted in terms of physical events responsible for the induction of the initial lesion. The quantitative relationship of these aberrations to the absorbed dose can be expressed in terms of simple mathematical models. The assessment of human exposure to ionizing radiation and other environmental mutagenic factors * Coordinated by the WHO Reference Centre for Chro- mosome Aberrations, Bureau of Human Ecology, Environ- mental Health Directorate, Ottawa, Canada. Requests for reprints should be addressed to the Director, WHO International Reference Centre, Bureau of Human Ecology, Environmental Health Centre, Tunney's Pasture, Ottawa, Ontario KIA OL2, Canada. 1 Director, Bureau of Human Ecology, Department of National Health and Welfare, Ottawa, Canada. ' Head, Radiation Cytogenetics Section, Human Develop- ment Division, Bureau of Human Ecology, Department of National Health and Welfare, Ottawa, Canada. 3Head, Environmental Mutagenic Factors Section, Bureau of Human Ecology, Department of National Health and Welfare, Ottawa, Canada. Present address: Soroka Medical Centre, Ben Gurion University of the Negev, Beer Sheva, Israel. ' Director, Statistical Research Service, Department of Agriculture, Ottawa, Canada. has been usefully approached in terms of quantifiable changes induced in the chromosomes of peripheral blood lymphocytes. For this purpose, the dose- response relationships for radiation-induced aberra- tions in the interphase chromosomes of the human peripheral blood lymphocyte, cultured in vitro, have been extensively studied during recent years. The relative consistency and accuracy shown by the relationship for exchanges as compared with other forms of relationship, e.g., blood cell count (1) and biochemical indicators (1-4), have made these aber- rations a preferred indicator of radiation exposure. Indeed, exchange frequency is being used as a biological indicator in several laboratories (5-11). The dose-exchange relationship is thought to be reasonably consistent within any given laboratory (e.g., 12, 13). However, as shown by Fig. 1 consider- able variability exists among different laboratories (12-32), particularly at low doses, which are of special concern for radiation safety. Consequently, the exposure estimates for a given exchange frequen- cy differ from laboratory to laboratory. To allow comparisons of data, the interlaboratory differences contributing to the observed variability must be resolved. Standardized procedures are essen- 3202 -373- J. D. ABBATT ET AL. 1.8 1.6 -J -J ul o 1.4 1.2 z ° 1.0 X 0.8 o 0.6 Z 0.4 0 0.2 0 Fig. 1. Yields of dicentric aberrations in peripheral blood leucocytes irradiated in vitro. The left side of the figure shows the results of culturing for 50-54 hours (-) or approximately 72 hours (---): Curve 1 represents the findings of Bajerska & Liniecki (15); 2, Schmid et al. (32); 3, Evans (20); 4, Norman & Sasaki (25); 5, Bender & Barcinski, (12); 6, Bora (9); 7, Norman et al. (24); 8, Mouriquand et al. (23); 9, Kelly & Brown (22); 10, Gooch et al. (21); 1 1, Bender & Gooch (14); 12, Visfeldt (30); 13, Van der Elst et al. (31); 14, pooled data from the present WHO study. The right-hand side of the figure shows individual data from the present study: 1, Canada; 2, Japan; 3, Denmark; 4, USSR; 5, United Kingdom; 6, Mexico; 7, USA (San Francisco); 8, USA (Nashville). tial for interlaboratory comparison of dose-response relationships and all other data. POSSIBLE REASONS FOR INTERLABORATORY SCORING DIFFERENCES It is conceivable that measurable differences exist among different laboratories in the assessment of aberration frequency. Prima facie, such differences may be attributed to any or all of the following four factors, among others: (1) the method of identification and analysis of chromosome aberrations; (2) differences in irradiation procedures, calibra- tion of radiation dose, and conditions of irradiation; (3) biological and cell culture variables, e.g., cul- ture duration, procedures, materials; and (4) possible differences in the response of lympho- cytes from different donors. The primary objective of the present study was to examine some aspects of the problems associated with the identification and analysis of chromosome aberrations, with a view to answering the following specific but interdependent questions. (1) Are there systematic differences among scorers in different laboratories in their selection of meta- phase chromosome spreads for scoring aberrations? (2) Do scorers recognize, classify, and enumerate chromosome aberrations in measurably different ways? (3) Are there recognizable and consistent differ- ences between scorers in their assessment of aberra- tion frequency? (4) If the answers to questions (1), (2), and (3) are in the affirmative, how do such differences influence the observed dose-effect relationships? (5) Do such factors as (a) the level of aberration frequency and (b) the quality of chromosome prep- 374 HUMAN CHROMOSOME ABERRATIONS aration have an influence on the quantitative as well as qualitative analysis of aberrations? EXPERIMENTAL DESIGN The experiment was designed primarily to provide reliable answers to the above questions. Valuable experience was gained from a pilot collaborative scoring study conducted during 1968-69 between the Human Cytogenetics Laboratory, Department of National Health and Welfare, Ottawa, Canada; the MRC Clinical and Cytogenetics Unit, Edinburgh, Scotland; and the Biology Division, Oak Ridge National Laboratory, Oak Ridge, USA. This study consisted of analysing a set of three slides, provided from Ottawa, for radiation-induced interphase chro- mosome aberrations in the human peripheral blood lymphocyte. Results from this limited study showed appreciable differences between scorers in the deter- mination of aberration frequency. The design of the main experiment has been described in detail elsewhere (4, 33). Essentially, it was planned with the view to accommodating 8 labo- ratories that initially expressed a desire to partici- pate. It consisted of two parts. The first part required evaluation of a set of precoordinated metaphase chromosome spreads, in which each scorer assessed a definite number of identical spreads. These were carefully selected to present a mixture of good and bad spreads and contained samples of all the common types of chromosome aberration. For this purpose two slides were prepared and sent to each scorer by the coordinating laboratory. This set of observations is referred to below as the " precoordinated study ". The second part of the study required evaluation of a series of slides by scoring cells selected freely from specified areas on the slides. In order to provide a wide range of aberration frequency, the cell cultures were irradiated with X-ray doses varying from 5 to 400 rads. There were 8 treatments altogether, including one control. Four slides from each treatment were selected for scoring so that every scorer was required to score a total of 32 slides. The balanced incomplete block design, with four repli- cates, ensured that after each participant had com- pleted study of the first slide of each batch of 4, a complete set of 8 doses had been analysed, and similarly for the second and subsequent slides. This set of observations is referred to below as the " free- scoring study ". In order to discern the possible influence of slide quality on scoring efficiency and to eliminate bias that could possibly arise from the scoring of slides of a quality specific to a particular laboratory, slides of graded quality (poor to good) were selected; scorers were asked to assess the quality of each slide. The experimental description (33) also included a scoring sheet previously agreed by all the participants to be suitable for purposes of the experiment. MATERIAL AND METHODS Freshly drawn human peripheral blood from a single healthy male donor with normal karyotype was used for the study. Culture of lymphocytes and cytological procedures Whole blood was cultured by a modification of the method of Hungerford (34). Cultures were set up immediately prior to irradiation. One hour following irradiation, PHA (Burroughs Wellcome) was added to the culture bottles in the proportion 0.01 ml of PHA to 1 ml of medium. The cultures were main- tained throughout the experiment at 37 'C. Fixations were made 52 h following irradiation. Dosimetry and irradiation of cultures The cultures were maintained at 37 °C before, during, and after irradiation. Irradiation was carried out with an X-ray machine operated at a peak of 250 kV and a current of 12 mA with the added filtration of 0.5 mm of copper and 1 mm of aluminium. The dose-rate was measured by an EIL portable electrometer, model 37A, with a 5-cm3 ionization chamber (type D37C). There was a total of 8 doses: 0, 5, 20, 50, 100, 200, 300, and 400 rads. The irradiation chamber was a 20-ml glass " medi- cine bottle ", the thickness of each side being approximately 2 mm, and the distance between sides through which the beam passed being 2 cm. The dose rate to the cells varied from 3.3 R/sec to 4.2 R/sec, depending on the distance of the cells from the wall of the irradiation chamber where the beam entered. The dose rate fell from 94% to 72% of the dose to the front of the chamber as the beam passed through the cell suspension. The dose received by the cells, calculated for the mid-portion of the chamber, was therefore within ±10% of the designated exposure. Procedures for the identification and interlaboratory scoring of chromosome aberrations Scoring procedures in one laboratory may differ from those in another. For instance, in recording 375 J. D. ABBATT ET AL. aberration types, some laboratories may prefer to record all, including gaps, whereas others may score only those aberrations relevant to the objectives, and those aberrations can be analysed with a high degree of accuracy. Furthermore, definitions of aberrations may vary among laboratories. In order to make the scoring as consistent and as comparable as possible, each of the participating laboratories was provided with a scoring guide containing general definitions of all the possible types of aberration (33). In order to facilitate rapid comparison of scoring by different laboratories, it was necessary to have uniform recording symbols and data entry proce- dures. For this purpose a mock scoring sheet with symbols and entries was sent to each of the scorers and each scorer was asked to enter the coordinates of each spread observed and scored. Two slides containing 118 selected spreads were chosen for the precoordinated study. Each of these spreads was prelocated with coordinates and each scorer had to relocate the spreads and to assess them for quality and aberrations. Two alternative pro- cedures for relocation of the precoordinated spreads were described in detail in the scoring guide book (33). One of the main steps in the procedure for scoring aberrations is the selection of suitable spreads. However, there is no standard method of spread selection and the decision to reject a spread can therefore be highly subjective. An attempt was made to classify rejected spreads into several categories and the scorers were asked to categorize the rejected spreads. In the free-scoring study, 32 slides, 4 for each dose, were chosen for scoring. The chromosome spreads for this study were not prelocated. However, in order to avoid possible variation of aberration frequency within a slide, the scorers were asked to confine scoring as far as possible to certain well-defined areas on the slides whose locations were described in the guide book. A minimum of 25 spreads had to be scored from the predefined area or areas on a slide. Statistical design for analysis of results The number of scorers participating in each of the 8 laboratories varied. In some, only one person was involved, in others two or more persons scored all the slides independently, while in yet others the slides were analysed jointly by 2 or more scorers. This created a number of problems in distinguishing block effect from error. The analysis therefore did not attempt to distinguish these two sources of error. The design is to be regarded as accommodating any trends over time in the scorers, such as increasing boredom or interest, deviation from specific scoring instructions, and possible deterioration in slide quality. If there is a perfect agreement between the scores recorded by each pair of the participants for each of the slides, the values recorded by one participant (i) plotted against those of another (j) will produce a straight line passing through the origin at an angle of 45°. This can be expressed as si = bsj (1) with b = 1 if there is perfect agreement. The analysis consists of estimating b and comparing the estimate with a true value of unity (35). Two response variables were studied in terms of this model-the exchanges per metaphase chromosome spread and the proportion of spreads with chromosome damage. Transmission of slides In the precoordinated study, the two slides to be scanned were circulated among the laboratories. They were first despatched on 4 May 1970 and finally received back on 20 February 1971, having been scored by the 8 laboratories. A 9th laboratory was included in December 1971. In the free-scoring study, the 32 slides were selected, coded, and placed in 8 packets of 4 slides each, according to a balanced incomplete block design with 4 replicates. Each packet was sent to each of the 8 participating laboratories with limited instructions on the prearranged transmission cycle. A series of notification sheets were used so that the transfer of the slides between different laboratories and countries could be followed by the coordinator and identified in case of postal delay or loss. The slides were first despatched on 1 May 1970 and received back on 11 May 1971. A total of 18 scorers from 9 laboratories partici- pated in the study. ANALYSIS OF INTERLABORATORY IDATA In the precoordinated study 118 preselected meta- phase spreads were scored from a set of 2 slides. For purposes of analysis it had to be assumed that all spreads had been properly located by all scorers. Thirteen scorers from 8 laboratories provided their individual assessments. However, it is possible that in those laboratories where two or more scorers participated some discussions may have taken place among the participants. 376 HUMAN CHROMOSOME ABERRATIONS 377 Table 1. Frequency of aberrations in 118 precoordinated metaphase spreads Classification of spreads Chromosome aberrations per spread scored Scorer code % total % % Exchanges a Acentric Deletions, Rejected scored Normal Damaged Deletions a Minutesa rings a minutes,acentric rings USSR 1 32 80 26 73 0.86 (69) 0.51 (41) 0.08 (6) 0.06 (5) 0.65 USSR 2 33 78 29 70 0.86 (67) 0.49 (38) 0.06 (5) 0.06 (5) 0.62 USA 2 53 54 27 72 0.83 (45) 0.20 (11) 0.35 (19) 0.11 (6) 0.67 USA 3 50 59 32 67 0.95 (56) 0.17 (10) 0.34 (20) 0.12 (7) 0.63 USA 4 60 47 27 72 1.00 (47) 0.17 (8) 0.32 (15) 0 (0) 0.49 Denmark 2 46 63 23 76 0.81 (51) 0.19 (12) 0.37 (23) 0 (0) 0.56 United Kingdom 2 55 53 41 58 0.68 (36) 0.11 (6) 0.45 (24) 0.06 (3) 0.62 United Kingdom 3 61 46 21 78 0.83 (38) 0.43 (20) 0.07 (3) 0.09 (4) 0.59 Mexico 3 85 17 11 88 0.82 (14) 0.24 (4) 0.12 (2) 0.06 (1) 0.41 Canada 1 23 90 21 78 0.96 (86) 0.31 (28) 0.11 (10) 0.14 (13) 0.57 Canada 2 35 76 19 80 1.09 (83) 0.36 (27) 0.11 (8) 0.17 (13) 0.63 Japan 1 38 73 17 82 0.93 (68) 0.12 (9) 0.27 (20) 0.07 (5) 0.47 mean = 0.98 mean = 0.64 a Numbers in parentheses denote number of aberrations observed. The results summarized in Table 1 show consider- able variation in the proportion of cells selected and scored. There is also no unanimous agreement among the 13 scorers even on one damaged spread. Table 2. Two-way table showing extent scoring the same precoordinated spreads Table 2 provides some indication of the range of scoring variability. Implications of these differences in calibrating the exposure dose will be discussed later. of agreement between two participants 01 02 03 04 05 06 07 08 09 10 11 12 13 o0 a 26 1 1 1 02 9 10 03 7 4 60 35 1 15 23 8 13 4 1 04 4 35 34 1 7 12 6 7 4 1 05 1 1 1 1 06 3 14 7 1 11 6 07 1 1 25 14 1 6 15 4 4 1 1 08 1 9 7 1 3 6 09 13 6 2 5 10 2 10 3 2 1 2 2 11 12 3 1 1 a 1. Rejected 2. Normal 3. Damaged 4. Damaged and dicentrics 5. Damaged and polycen- trics 6. Damaged and centric ring 7. Damaged and deletion 8. Damaged and minutes 9. Damaged and acentric- ring 10. Damaged and single gap 11. Damaged and double gap 12. Damaged and break 13. Damaged and exchange. J. D. ABBATT ET AL. Table 3. Frequency of aberrations in laboratory the free-scoring study: pooled data for each Country Dose ~Total cells Total %Dmgd Ecags DltosCountry scored damaged cells per cell minutes,cells ~~~~~~~~acentric rings Canada 0 215 2 0 0 0.00 5 216 7 3 0.019 0.01 20 220 13 5 0.018 0.04 50 205 23 11 0.07 0.03 100 210 60 28 0.20 0.12 200 206 131 63 0.70 0.26 300 201 174 86 1.29 0.63 400 203 189 93 1.87 0.85 Japan 0 143 2 1 0 0.01 5 141 2 1 0.007 0.04 20 147 5 3 0.02 0.01 50 128 13 10 0.055 0.02 100 150 43 28 0.20 0.09 200 139 83 59 0.59 0.30 300 146 133 91 1.26 0.49 400 134 128 95 1.93 0.54 Denmark 0 100 2 2 0 0.01 5 100 3 3 0.01 0.01 20 101 6 5 0 0.03 50 101 9 8 0.03 0.03 100 103 27 26 0.15 0.08 200 90 50 55 0.48 0.19 300 100 87 87 1.04 0.49 400 105 98 93 1.50 0.79 USSR 0 203 6 2 0.004 0.02 5 205 12 5 0.020 0.02 20 202 10 4 0.025 0.01 50 201 24 11 0.045 0.06 100 206 58 28 0.180 0.13 200 200 120 60 0.515 0.29 300 200 154 77 1.00 0.52 400 206 193 93 1.665 0.96 United Kingdom 0 343 7 2 0 0.01 5 335 3 0 0.006 0.00 20 410 17 4 0.017 0.02 50 204 11 5 0.02 0.04 100 115 27 23 0.17 0.10 200 149 74 49 0.47 0.22 300 209 157 75 0.97 0.48 400 168 153 91 1.63 0.72 Mexico 0 131 4 3 0.08 0.02 5 100 5 5 0.010 0 20 130 5 3 0.008 0.01 50 82 12 14 0.061 0.05 100 74 16 21 0.189 0.07 200 84 48 57 0.33 0.37 300 62 29 46 0.629 0.08 400 58 52 89 1.655 0.55 USA 0 224 3 1 0 0.01 (San Francisco) 5 201 5 2 0 0.02 20 225 8 3 0.027 0.00 50 200 10 5 0.010 0.04 100 228 49 21 0.180 0.08 200 201 114 56 0.562 0.24 300 224 168 75 1.116 0.47 400 199 179 89 1.784 0.75 USA 0 182 5 2 0 0 (Nashville) 5 212 10 4 0.005 0.01 20 203 14 6 0.019 0.03 50 156 22 14 0.032 0.05 100 113 31 27 0.21 0.09 200 124 68 54 0.44 0.19 300 189 156 82 1.12 0.47 400 175 168 96 1.79 0.74 378 HUMAN CHROMOSOME ABERRATIONS Table 4. Chromosome exchanges per metaphase spread found by different scorers in the free-scoring study Scorer code (R) USSR USSR USA USA USA USA Den- Den- Mex- Mex- Mex- Mex- Can- Can- J(R) U S mark mark~UK UK ico ico ico ico ada adaJan1 2 1 2 3 1 2 1 2 1 2 3 4 1 2 1 0 0 0.01 0 0 0 0 0 0 -a 0 0 - 0 0.012 0 0 0 5 0.01 0.03 - 0 0 0.009 0 0.019 0 0.008 0 - - 0.023 0.02 0.17 0.007 20 0.02 0.03 0 0.03 0.03 0.0197 0 0 - 0.017 0 - 0 0.02 0.03 0.008 0.02 50 0.04 0.05 - 0 0.02 0.032 0 0.06 - 0.02 - - 0.083 0.043 0.08 0.067 0.055 100 0.19 0.17 0.20 0.17 0.184 0.212 0.167 0.08 - 0.174 0.14 - 0.20 0.20 0.22 0.18 0.2 200 0.53 0.50 - 0.54 0.58 0.44 0.41 0.50 - 0.47 0.26 - - 0.39 0.79 0.61 0.59 300 0.95 1.05 1.29 1.08 1.11 1.07 0.79 1.12 - 0.97 0.93 - - 0.36 1.36 1.22 1.26 400 1.56 1.78 - 1.68 1.89 1.79 0.92 1.69 - 1.63 1.80 1.36 - 1.70 1.86 1.92 1.93 a Slide not scored. A total of 17 scorers participated in the free- scoring study. However, only 13 scorers completed analysis of the 32 slides. Although each scorer was asked to score only 25 spreads from a defined area or areas of a given slide, i.e., 100 spreads per dose, the actual number of spreads scored per dose varied from about 60 to 400. Intralaboratory and interlaboratory variability in scor- ing chromosome exchanges Calibration coefficients. The results are sum- marized in Tables 3 and 4. In fitting the calibration expression (equation 1), advantage was taken of the quadruplicate replication of each dose; it was found that the relationship that existed between the mean and the variance for each dose was removed by using the square root of the exchange frequency per cell. The estimates of b in equation 1 for the square root of the exchange per spread for each pair of scorers are presented in Table 5. Of 85 values computed, only 23 did not differ significantly from unity. Because of this large disagreement among scorers, a rough calibration of each participating scorer was made against the majority, that is, the pooled remainder; only 3 scorers agreed with the general trend. Relationship between exchanges per spread and dose. The dose-response data for human chromo- somes obtained by different laboratories during the past decade have so far been applied to one or more of the three possible models, namely: y = c' + aD y = c + PD2 y = c"' + a'D + fl'D2 (2) (3) (4) where y is the number of exchanges per spread; D, the radiation dose in rads; a and fi, the coefficients of one-hit and two-hit aberrations; and c, the back- ground frequency of exchanges. Most data have been fitted to the quadratic model (equation 4), based on the original hypothesis by Lea & Catcheside (36) that a proportion of the exchanges can result from a single hit, and that the relative proportion of such single-hit exchanges rapidly diminishes as the dose is increased. The present data have been fitted to the quadratic model, slightly modified, as follows: k y=c+aD+lD2+ £ bjzj j=I (5) where y, D, a, and ,B are as before, zj are k design variables for laboratories and for scorers within laboratories, and bj denotes parameter estimates for the coefficients zj. Since the number of spreads scored by different scorers varied from 25 to 125 per slide, a weighted analysis was performed, the weights being the num- ber of spreads examined. The results of an analysis performed by the method of fitting constants are summarized in Table 6. The data from each labora- tory were fitted to equation 4 and it was found that in no case did c differ significantly from 0; consequent- ly we fitted y = aD + [3D2 (equation 4); the values 379 J. D. ABBATr ET AL. Table 5. Estimates of calibration coefficients among scorers in the free-scoring study: exchanges per spread- square-root transformation to remove relationship between variance and mean USSR USSR USA USA USA Den- Den- UK Mex- Mex- Mex- Can- Can- 1 2 2 3 4 mark mark 2 ico ico ico ada ada1 2 1 2 3 1 2 USSR 2 1.032 USA 2 1.016 0.984 USA 3 1.058 1.025 1.041 USA 4 1.359 1.319 1.333 1.281 Denmark 1 0.845 0.816 0.878 0.778 0.662 Denmark 2 1.018 0.987 0.985 0.972 0.736 United Kingdom 2 1.313 1.272 1.291 1.241 0.970 1.609 1.272 Mexico 1 0.643 0.719 0.659 0.627 0.507 0.668 0.534 Mexico 3 0.938 1.430 0.934 0.930 0.594 0.817 Mexico 4 0.742 0.655 0.730 0.702 0.545 0.687 0.827 0.522 Canada 1 1.139 1.103 1.121 1.076 0.839 1.363 1.115 0.867 1.761 1.233 1.575 Canada 2 1.098 1.064 1.081 1.038 0.810 1.304 1.077 0.836 1.599 0.857 1.548 0.965 Japan 1 1.295 1.255 1.276 1.222 0.957 1.577 1.255 0.988 1.815 1.448 1.876 1.137 1.179 The values in italics indicate relationships significantly different (P > 0.05) from unity. Blanks indicate too few data for a calibration estimate; 3 scorers (USA 1, United Kingdom 1, Mexico 2) did not provide sufficient data for calibration and/or significance tests. Table 6. Analysis of variance of exchanges per spread in the free-scoring study: variability in the quadratic model among laboratories and among scorers within laboratories where there were two or more. Source of variability DF Mean square Regression on dose: Linear term 1 3525.2201 a Additional due to quadratic term 1 100.2227 a Between laboratories 7 2.5708 a Between scorers within laboratories 9 0.5897 Residual 340 0.8718 a p < 0.01 of a and ,B so obtained are presented in Table 7. It is clear from the analysis of variance (Table 6) that the contributions made by the quadratic term to the dose-response relationship are significant. Further- more, it is evident that there are significant differ- ences among laboratories in the estimates of a and , the former varying from 0.09 x 10-3 to 2.00 x 10-3 and the latter from 7.03 x 10-6 to 9.92 x 10-6. The respective pooled values for all the laboratories are a = 1.05 x 10-3 and , = 8.33 x 10-6. However, the differences within a laboratory, where known, were only once on the borderline of significance. The data on the frequency of dicentrics from this study were fitted to the quadratic expression (equa- tion 4). The values of single-hit (a) and two-hit coefficients (,B) ranged from 0.41 x 10-3 to 2.47 x 10-3 and 2.95 x 10-6 to 9.55 x 10-6 respectively. The pooled dicentric data for all the laboratories can be expressed as follows: y = 1.16 x 10-3D + 6.09x 10-6D2 The dose-response curve obtained for dicentrics can be compared with other curves obtained prior to this study by some of the participating laboratories and others in Figure 1. It may be noted that the spread in the curves in the left-hand half of Figure 1 is much greater than in the right-hand half. Whereas, in the present study, the estimates of exchanges per cell among all laboratories, except one, vary by a factor of only about 1.5, in previous studies it varied by a factor of the order of 5 or more. Intralaboratory and interlaboratory variability in scor- ing deletions and other chromosome aberrations For the purposes of this report it was considered that a deletion is probably a one-hit aberration, 380 HUMAN CHROMOSOME ABERRATIONS Table 7. Coefficients of exchanges (quadratic model) in the free-scoring study a Laboratory N a x 10-3 sxe X 1-5 s.1 b R c x3 X 10-3 X10-5 USSR (Moscow) 64 .913 .413 .813 .121 .975 USA (San Francisco) 68 1.005 .288 .876 .087 .989 USA (Nashville) 32 .622 .431 .966 .126 .988 Denmark (Copenhagen) 32 1.040 .579 .703 .169 .973 United Kingdom (Harwell) 32 .734 .537 .832 .157 .978 Mexico (Mexico City) 35 .090 1.052 .992 .311 .908 Canada (Ottawa) 64 1.999 .343 .694 .100 .988 Japan (Tokyo) 32 1.358 .558 .871 .163 .984 Pooled data 359 1.050 .188 .833 .055 .973 a Number of slides scored. b Standard error. c Multiple correlation coefficient. whereas an acentric ring and a pair of minutes are probably two-hit aberrations. However, it is often difficult to distinguish one type of aberration from the other and therefore the uncertainty in scoring these aberrations is much greater than that in scoring exchanges. For this reason, attempts are seldom made to relate these aberrations, either individually or collectively, to dose and to interpret the relation- ships in terms of physical and biological events. However, some indications of the magnitude of variation in the ensuing results can been seen in Tables 1 and 3. No attempt is made here to provide a model for these relationships or to evaluate the extent of interlaboratory scoring variability, for the reasons stated above. It is of interest to note that the enormous differences seen in the scoring of individ- ual aberrations almost disappear when the three types are combined (Table 1). This observation clearly suggests that the problem of scoring these aberrations is connected not so much with the recognition of the damage as with their classifica- tion. Relationship between the dose and the frequency of cells with chromosome damage For the purpose of calibration (see equation 1), a spread is defined as damaged if it contains any chromosome aberration or a combination of chro- mosome aberrations, including symmetrical translo- cations. Since the proportions of such cells, x, cannot be expected to be normally distributed, the angular transformation of x, as defined by: y = sin-1x (6) was used for computations. The calibration coeffi- cients and the actual proportions of the damaged spreads are shown in Table 8, where it can be seen that, as compared with the coefficients of exchange frequencies, there are relatively few disagreements, 64 values not differing significantly from unity. Probit analysis In a population of cells that has been exposed to a radiation dose within the range under study, some cells will have some form of visible chromosome anomaly and others will have none. Since the physical events that lead to aberrations occur at random and are localized, the response of one cell is considered to be independent of that in others. Therefore, the probability that either type of cell will occur in an irradiated cell population can be consid- ered binomial, and the data can be examined in terms of the probit model (37). Three analyses were performed, by fitting a line for each scorer, fitting parallel lines for scorers in a laboratory, and fitting one line to the scorers at a laboratory. Likelihood ratio tests showed that scor- ers from different laboratories and different laborato- ries themselves differed significantly in scoring damaged spreads (Table 9). In two cases the differ- ences within laboratories were significant. 381 382 J. D. ABBATr ET AL. Table 8. Calibration coefficients among scorers in the free-scoring study: percentage of damaged spreads-angular transformation USSR USSR USA USA USA Den- Den- UK Mex- Mex- Mex- Can- Can- 1 2 2 3 4 mark mark 2 ico ico ico ada ada1 2 1 2 3 1 2 USSR 2 USA 2 USA 3 USA 4 Denmark 1 Denmark 2 United Kingdom 2 Mexico 1 Mexico 3 Mexico 4 Canada 1 Canada 2 Japan 1 1.053 0.946 0.899 0.973 0.924 1.054 1.002 1.014 0.888 1.061 1.044 0.960 0.918 0.904 0.902 1.178 1.159 1.034 0.952 1.127 1.070 1.022 0.971 1.101 1.045 1.028 1.114 1.078 1.119 1.015 0.935 1.405 1.122 1.190 1.080 1.163 1.084 1.008 1.105 0.986 0.951 1.190 1.062 1.159 1.051 1.130 0.992 0.996 0.913 0.852 0.941 1.018 1.068 0.970 1.043 0.931 0.911 0.833 0.972 1.316 0.976 1.056 1.074 1.157 1.044 1.171 1.064 0.942 1.064 1.111 1.027 1.146 1.248 1.075 1.138 1.116 0.730 1.021 0.908 1.224 0.977 1.058 0.977 1.076 The values in italics indicate relationships significantly different (P > 0.05) from unity. Blanks indicate too few data for a calibration estimate; 3 scorers (USA 1, United Kingdom 1, Mexico 2) did not provide sufficient data for calibration and/or significance tests. Table 9. Free-scoring study: summary of results of fitting probit lines to numbers of damaged spreads as a function of number examined and dose Log likelihood Laboratory No. ofscorers Separate Chi 2, Parallel --Chi 2. One linelines lines USSR (Moscow) 2 -113.5 0.28 -113.7 1.90 -114.6 USA (San Francisco) 3 -112.3 4.06 -114.4 0.76 -114.7 USA (Nashville) 1 - . -80.6 Denmark (Copenhagen) 2 -51.8 26.51 a -65.0 - -66.9 United Kingdom (Harwell) 2 -61.2 0.01 -61.2 0.00 -61.2 Mexico (Mexico City) 4 -47.7 0.18 -47.8 3.99 -49.8 Canada (Ottawa) 2 -137.2 8.93 a -141.7 - -147.1 Japan (Tokyo) 1 - - -63.4 All scorers 17 -667.8 66.33 a -700.9 _ -733.3 Laboratories 8 -698.4 *-Chi 2 _69.77 a -733.31 - a Significant likelihood ratio test (P < .05). HUMAN CHROMOSOME ABERRATIONS Table 10. Free-scoring study: multivariate analysis of variance of number of damaged spreads and number of exchanges among variables as indicated Predictor variables Hypothesis X 2 DF Model 1 A. Number of spreads scored Model 1 is not relevant to the data 832.5 a 48 B. Between laboratories C. Scorers/laboratories D. Doses (dummy variables) Model 2 A. Number of spreads scored (a) Model 2 is not relevant to 757.4 a 16 the data D. Doses (dummy variables) (b) Model 1 does not differ from 85.7 a 32 model 2 a p < 0.001. Multivariate analysis of variance In this analysis, both the number of damaged spreads and the number of exchanges were analysed simultaneously. The model can be written: 7 k y1= ao, + all x + v bm1dm + E Cjl zj (7a) m=rI j=1 7 k Y2 = aO2+ a12 x++X bm2 dm + l; cj zj (7b) m=1 j=l where Yi is the number of damaged spreads, Y2 is the number of exchanges, x is the number of spreads examined, bm1, bM2 and dm are the dose levels expressed as 7 design variables, and zj (j = 1, 2, k) are appropriate design variables for the scorers and laboratories. The X2 computed from U statistic to examine some hypotheses are given in Table 10. A comparison of model 1 and its reduced form, model 2, which omits differences among laboratories and scorers, reveals that there is hetero- geneity among scorers, that is, the differences that exist are unlikely to have occurred by chance. :DISCUSSION The results from the precoordinated study, as well as those from the free-scoring study, clearly indicate that systematic differences exist among different laboratories in the scoring of chromosome exchanges and therefore in the estimates of the coefficients in the quadratic dose calibration expression. The differ- ences were larger among individual scorers from different laboratories than among scorers in the same laboratory and were more noticeable in the free- scoring than in the precoordinated spreads. This may suggest a more cautious approach (bias) on the part of the scorers towards scoring the preselected spreads. They may have been conscious of the fact that the individual assessment of each of the prese- lected spreads was going to be compared with others. The precoordinated study revealed considerable differences among scorers in the recognition and classification, and therefore in the enumeration, of different types of aberrations, especially deletions, acentric rings, and minutes. Discernible differences were also apparent in the selection of spreads for scoring. In some laboratories, cells were rejected at a much higher rate than in others, e.g., with slide 9, 24 of a total of 30 spreads were rejected in one laboratory as compared with a mean of about 12 for all the laboratories taken together. The rate of rejection of spreads would not be expected to have any significance in the yield of aberrations per scored spread if rejection is random. The linearly inverse relationship between the total yield of exchanges and the number of cells observed in the precoordinated study tended to confirm this expectation. To investi- gate this possibility further, the data from those individuals who participated in both studies were selected. When they were analysed, the number of exchanges per spread, the three expressed opinions on slide condition (i.e., quality, frequency, and staining), and the dose from the free-scoring study were considered. From the precoordinated data, the rejection proportion was used; these proportions are linked, since they are the various opinions on the same 118 spreads. An analysis of variance in Table 11 suggests the following conclusions: (1) The opinion of scorers as to the slide condi- tion, i.e., quality (Q), frequency (F), and staining (S), which are hedonic data, showed no significant association with the observed variability of exchange rate among individuals, among dose levels, or among their interactions (Tables 6, 7, and 11). 383 J. D. ABBATT ET AL. Table 11. Analysis of variance li in precoordinated study to exc dose, and slide quality in free-sco Source of variability Among individuals Rejection Quality (Q), frequency (F), staining (S) Residual Among dose levels regression on dose Linear Quadratic Deviations from quadratic Q, F, and S Residual ndividuals x dose levels Regression x individuals Regression x rejection Residual Q, F, and S Residual a 0.01 < P 6 0.05 b 0.001 < P 6 0.01 c p < 0.001 nking rejection rates another might be a combination of poor spread :hanges per spread, quality and heavy chromosome damage. The latter ,ring study possibility is supported by the fact that the laborato- DF Mean square ries that rejected fewer numbers of cells usually scored a larger number of damaged spreads. The large differences in the dose-effect relation- 1 0.1401 b ships for dicentrics among laboratories in the pre- vious studies as compared with those in the present 3 0.0049 study strongly suggest that factors other than differ- 6 0.0140 ences in scoring aberrations are involved. Among these, culture and irradiation procedures may play 1 30.5906 c important roles.It is clear from this study that real differences may 1 1.4114 c exist not only among individual scorers but also 5 0.0019 among and within different laboratories, a situation 3 0.0015 that must be of serious concern to all those using 2 0.0024 chromosome damage as a biological dosimeter andthat emphasizes the need for individual " calibra- tion " of each laboratory group. 20 0.0205 c The degree of agreement among scorers appears to 2 0.1045 b be higher for damaged spreads than for exchanges. 18 0.0112 a This finding, together with the fact that at low doses 3 0.0072 aberrations other than exchanges contribute substan-0*0072 tially to the frequency of damaged cells, suggests that 47 0.0060 the frequency of damaged cells may be a better indicator of radiation damage than the frequency of exchanges. However, much work needs to be done before cell damage can be considered as a possible index for biological dosimetry. The following may be the reasons for the different degrees of agreement:(2) The rejection rate accounted for some of the differences among the regressions, both of shape and position. This situation may be summarized as a series of correlation coefficients; the simple correlation be- tween exchanges per spread and rejection rate with 85 degrees of freedom was -0.065, the partial correlation after allowing for the regression was -0.381 with 83 degrees of freedom, and the partial correlation after allowing for the regression and slide condition was -0.370 with 80 degrees of freedom. The apparent lack of relevance of slide condition, in spite of the real differences in opinion which existed (Table 11), is unexpected and somewhat encourag- ing. The reasons for the large differences in cell selection in the present study are not clearly evident. Although the assessment of slide quality varied among the laboratories, there was no clear correla- tion between poor slide quality and high rate of rejection of spreads. An individual preference for a certain type of spread might be one reason, while (1) A real difference may exist between the scoring of chromosome aberrations and damaged cells. It is quite likely that many of the spreads that are rejected as poor-quality spreads for scoring aberrations will otherwise be scored as damaged. (2) It has taken over a decade and required extensive investigations to describe the dose-ex- change relationship in terms of mathematical models that can be interpreted in terms of physical and biological events and are generally acceptable. Simi- lar efforts have not yet been made with regard to relating damaged cells to the dose, although Brown and his collaborators (3840) have been using the proportions of damaged cells as a quantitative measure of radiation damage during the past few years. A factor favouring the method is the known wide applicability of the probit model for dose- response studies (37). Any consideration of damaged cells as a biological indicator must be preceded by further quantitative studies on dose-effect relation- ships in this context. 384 HUMAN CHROMOSOME ABERRATIONS It is apparent from this study that, in terms of either chromosome aberrations or damaged cells, considerable assessment differences exist within the laboratories in the qualitative as well as quantitative measurement of radiation damage and therefore in providing quantitative estimates of radiation expo- sure. A uniform and generally acceptable method of biological dosimetry of radiation exposure can be developed only through collaborative efforts of individual cytogeneticists and the laboratories concemed. SUMMARY AND) CONCLUSIONS This WHO-sponsored intemational study was conducted with a view to identifying possible causes for the large interlaboratory differences in assessing the relationship between chromosome aberrations and absorbed radiation doses in the human peripher- al blood lymphocyte. A total of 18 scorers from 9 laboratories in 7 countries participated in the study, which consisted of two parts and involved scoring interphase chromosome aberrations. In the first part of the study, a number of precoordinated metaphase spreads from a set of 2 slides were analysed by scorers from all laboratories; in the second part, a group of 32 slides from blood samples irradiated in vitro with X-ray doses varying from 5 to 400 rads at 3.45 rads per min were scored by scorers from the same laboratories, except in one case. The first part of the study revealed significant differences among the laboratories, not only in the recognition and classification of various types of aberrations but also in the selection of spreads for scoring. The data on chromosome exchanges from the second part showed, in general, a better fit to a quadratic model, y = aD + flD2 (where y is the exchanges per cell, D the radiation dose, and a and ,B the one-hit and two-hit components) than to either a linear or a dose-squared model. Large and significant variation was observed among laboratories in the assessment of a and ,B for exchanges, the former varying from 0.09 x 10-3 to 2.00 x 10-3, and the latter from 7.03 x 10-6 to 9.92 x 10-6. The corresponding values for dicentrics were 0.41 x 10-3 to 2.47 x 10-3, and 2.95 x 10-6 to 9.55 x 10-6. The pooled data on dicentrics alone and all two-hit aberrations taken together can be expressed as y = 1.16 x 10-3D + 6.09 x 10-6D2, and 1.00 x 10-3D + 8.50 x 10-6 D2, respectively. Variation in the frequency of damaged spreads was also significant among the laboratories. How- ever, it was smaller than that for exchanges. The variation in the dose-response relationship in the present study appears to be much smaller than that in the relationships previously reported by several of the participating laboratories and others. It is evident from this study that there were appreciable differ- ences among scorers within and among laboratories in the selection of spreads for scoring, and also in the recognition and classification of different types of aberrations. There is a correlation between these differences and the aberration frequency. This correlation may be connected with the degree of experience of the scorers involved, and provides an obvious explana- tion for observed differences. However, it has not been possible to relate these differences quantitative- ly to the differences in the aberration yields. It can be inferred that factors other than scoring differences must play important roles in determining the large variation seen in the published dose-response data. It is concluded that significant interlaboratory variability exists in the analysis of human metaphase chromosome spreads for radiation-induced inter- phase aberrations. This appears to be related partly to possible differences in the selection and rejection of spreads for scoring and also in the recognition and classification of various types of aberrations. It is reflected in the dose-response relationships for aber- rations and as well as for damaged spreads. How- ever, the scoring variability appears to be relatively small for the latter. It can be inferred that factors other than differences in scoring aberrations also play a significant role in the large variation in the dose-response relationships seen in the published literature. SUGGESTED FUTURE INVESTIGATIONS This study has clearly shown that significant differences exist among 9 leading cytogenetic laborat- ories in the analysis of various types of interphase chromosome anomalies in the human peripheral blood lymphocyte. Although it is possible to relate an unascertained part of this difference to detectable differences in the selection of cells for analysis and in the recognition and classification of aberrations, a quantitative correlation cannot at present be estab- lished between these differences. There is evidence that biological and culture variables such as the length of lymphocyte culture time, differences in the radiation response of individ- ual donors, variable environmental factors (e.g., temperature and oxygen level), and differences in 385 J. D. ABBATT ET AL. physical parameters (e.g., quality of radiation, range of dose, dose-rate, dose fractionation, and dosimetric procedures), individually or in combination, can have a considerable influence on the dose-response relationships. Further interlaboratory collaborative study is therefore required to investigate the role of these factors. As it is not practicable to study all the factors simultaneously in one study, it will be necessary to decide on the order in which they could be studied. It may be useful to organize a workshop- type discussion among the cytogeneticists concerned to formulate a comprehensive approach to identify the causes of interlaboratory differences and to relate these to the variation in the dose-response relation- ship. * * * PARTICIPATING LABORATORIES Institute of General Genetics USSR Academy of Sciences Moscow, USSR (Professor N. P. Dubinin) Atomic Energy Authority Health and Safety Branch Radiological Protection Division Harwell, England (Dr G. W. Dolphin) Institute of Pathology Frederiksberg Hospital Copenhagen, Denmark (Dr J. Visfeldt) MRC Clinical and Population Cytogenetics Unit Western General Hospital Edinburgh, Scotland (Miss Karin E. Buckton) Instituto Nacional de Energia Nuclear, Radiobiologia y Genetica Mexico 18, D.F. (Dr Alfonso Le6n de Garay) San Francisco Medical Center University of California San Francisco, Cal., USA (Dr S. Wolff) Department of Human Cytogenetics Tokyo Medical and Dental University Tokyo, Japan (Dr Masao S. Sasaki) The Johns Hopkins University School of Hygiene and Public Health Department of Radiological Science Baltimore, Md., USA (Dr Michael A. Bender) Statistical Research Services Department of Agriculture Ottawa, Ontario, Canada (Mr L. P. Lefkovitch) Bureau of Human Ecology Environmental Health Protectorate Health Protection Branch Department of National Health and Welfare Ottawa, Ontario, Canada (Dr John D. Abbatt) ACKNOWLEDGEMENTS The coordinators and collaborating laboratory workers responsible for this study thank Dr W. Seelentag and Dr E. I. Komarov of the World Health Organization for their advice and help. They also gratefully acknowledge the helpful criticism provided by a large number of scientists, in particular Dr I. R. Le Go, Mme Doloy, Dr I. M. Bauchinger, Dr J. Liniecki, and Dr I. N. P. Bochkov. RPSUM1 ETUDE INTERNATIONALE DE REFERENCE SUR L'IDENTIFICATION ET LA NOTATION DES ABERRATIONS CHROMOSOMIQUES HUMAINES: RESULTATS D'UNE ETUDE COMPARATIVE DE L'OMS Les aberrations chromosomiques sont des parametres reconnus des lesions cellulaires provoqu6es par les rayon- nements ionisants ou par certaines substances chimiques. Elles peuvent ais6ment faire l'objet d'une evaluation quantitative et on estime que dans un laboratoire entrain6 cette evaluation fournit des donnees relativement cons- 386 HUMAN CHROMOSOME ABERRATIONS 387 tantes. Cependant, il reste a etablir si les resultats obtenus par differents laboratoires sont comparables et a eliminer les facteurs qui sont a l'origine de divergences en faisant notamment appel a des techniques normalisees pour la mesure des rapports dose d'irradiation-reponse. Le pr6sent article expose les resultats d'une etude internationale, patronn6e par l'OMS, ayant pour objet d'identifier les causes possibles des fortes variations observ&es d'un laboratoire a l'autre dans l'appreciation des rapports entre l'exposition aux rayonnements ioni- sants et l'apparition d'aberrations chromosomiques dans les lymphocytes du sang peripherique chez l'homme. Dix-huit examinateurs de 9 laboratoires, dans 7 pays, ont participe a cette etude qui comportait deux aspects. Dans un premier temps, les examinateurs de tous les laboratoires ont eu a analyser un nombre donne de cel- lules en metaphase presentant des aberrations, et prea- lablement selectionnees; dans un deuxieme temps, chaque exaniinateur a du etudier 32 preparations obtenues a partir de cultures cellulaires traitees par les rayons X a la dose de 5 a 400 rads. On a enregistre de notables differences, selon les labo- ratoires, non seulement dans l'identification et la classi- fication des divers types d'aberrations, mais aussi dans le choix des cellules a analyser. On a aussi not6, dans une moindre mesure, des variations entre laboratoires dans l'evaluation des cellules endommag6es. Dans la pr6sente etude, les divergences concemant les rapports dose d'irradiation-reponse sont apparues beau- coup plus faibles que celles rapportees anterieurement par divers laboratoires. De toute evidence, les apprecia- tions concernant le choix des cellules et l'identification et la classification des aberrations different de maniere sensible parmi les examinateurs d'un meme laboratoire et entre examinateurs de plusieurs laboratoires. II existe entre ces differences et la frequences des aberrations chro- mosomiques une correlation que l'on peut rapporter au niveau d'experience des examinateurs concemrs; nean- moins, il est certain que d'autres facteurs sont egalement a l'origine des variations importantes constatees dans les donnees publiees concemant les rapports dose d'irra- diation-reponse. REFERENCES 1. BOND, V. P. ET AL. Mammalian radiation lethality, New York, Academic Press, 1965. 2. SmTrH, H. & BATES, T. H. Dosimetry for radiation accidents, Vienna, International Atomic Energy Agency, 1965, p. 199. 3. GERBER, G. B. Accidental irradiation at work, Euratom, 1967, p. 141. 4. ABBATT, J. D. Biochemical indicators of radiation injury in man, Vienna, International Atomic Energy Agency, 1971. 5. BENDER, M. A. Ann. N. Y. Acad. Sci., 114: 249 (1964). 6. BENDER, M. A. & GOOCH, P. C. Radiat. Res., 29: 568 (1966). 7. BUCKTON, K. E. & PIKE, M. C. Nature (Lond.), 202: 714 (1964). 8. BREWEN, J. G. ET AL. Radiat. Res., 49: 647 (1972). 9. BoRA, K. C. Genetics, 74: 527 (1973). 10. LLoYD, D. C. ET AL. Nature (Lond.), 241: 69 (1973). 11. QUASTEL, M. R. ET AL. Canad. Med. Ass. J., 102: 369 (1970). 12. BENDER, M. A. & BARaCNSKi, M. A. Cytogenetics, 8: 241 (1969). 13. EVANS, H. J. Proc. roy. Soc. B, 70: 132 (1968). 14. BENDER, M. A. & GOOCH, P. C. Proc. nat. Acad. Sci. (Wash.), 48: 522 (1962). 15. BAJERsKA, A. & LNmcKi, J. Int. J. Radiat. Biol., 16: 467 (1969). 16. BoRA, K. C. & SOPER, L. Canad. J. Genet. Cytol., 13: 364 (1971). 17. BREWEN, J. G. & LuIPoLD, H. D. Mutat. Res., 12: 305 (1971). 18. BAUCHINGER, M. In: Hug, 0. & Uppinger, A., ed. Sonderbuch aus Handbuch der medizinischen Ra- diologie, Berlin, Springer, 1972, p. 127. 19. DuBmiN, N. P. In: Problems of radiation genetics, Washington, D.C., United States Atomic Energy Control, 1962, p. 371 (Tr. 5376). 20. EVANS, H. J. In: Evans, H. J. et al., ed. Human radiation cytogenetics, Amsterdam, North Holland, 1967, p. 20. 21. GOOCH, P. C. ET AL. Biological effects of neutron and proton irradiations, Vienna, International Atomic Energy Agency, 1964, p. 325. 22. KELLY, S. & BROWN, C. D. Amer. J. pubi. Hlth., 55: 1419 (1965). 23. MOURIQUAND, C. ET AL. Report CEA-R 3007, Paris, Commissariat de l'Energie atomique, 1966. 24. NORMAN, A. ET AL. Radiology, 83: 108 (1964). 25. NORMAN, A. & SASAKI, M. S. Int. J. Radiat. Biol., 11: 321 (1966). 26. UNITED NATIONS SciENTrIFIc COMMrIrEE ON THE EFFECTS OF ATOMIC RADIATION. Radiation-induced chromosome aberrations in human cells, United Nations, New York, 1969 (A/AC 82 IR 231). 27. SCHMICKEL, R. Amer. J. Hum. Genet., 19: 1 (1967). 28. Scorr, D. ET AL. Mutat. Res., 8: 367; 9: 225 (1969). 388 J. D. ABBATr ET AL. 29. WOLFF, S. Mutat. Res., 15: 435 (1972). 30. VISFELDT, J. Riso Report 117, Riso, Danish AEC Research Establishment (1966). 31. VAN DER ELST, E. ET AL. Report of Comisi6n Nacional de Energia At6mica, Buenos Aires (1967). 32. ScHMm, E. ET AL. Mutat. Res., 16: 307 (1972). 33. BoRA, K. C. ET AL. Addendum to: Abbatt, J. D. Biochemical indicators of radiation injury in man, Vienna, International Atomic Energy Agency, 1971, p. 164. 34. HUNGERFORD, D. A. Stain Techn., 40: 333 (1965). 35. WILLiAMS, E. J. Regression analysis, New York, Wiley, 1959. 36. LEA, D. G. & Catcheside, D. G. J. Genet., 44: 216 (1942). 37. FINNEY, D. J. Probit Analysis, Cambridge, Cam- bridge University Press, 1962. 38. BUCKTON, K. E. ET AL. Lancet, 11: 676 (1962). 39. BUCKTON, K. E. ET AL. In: Evans, H. J. et al., ed. Human radiation cytogenetics, Amsterdam, North Holland, 1967, p. 122. 40. COURT-BROWN, W. M. ET AL. Lancet, 1: 1239 (1965).