285
Mutation Research, 50 (1978) 285--291 © Elsevier/North-Holland Biomedical Press
A CRITIQUE OF THE COLLABORATIVE CYTOGENETICS STUDY TO MEASURE AND MINIMIZE INTERLABORATORY VARIATION
CARROL
S. W E I L
Carnegie-Mellon Institute of Research, Carnegie-Mellon University, 4400 Fifth Avenue, Pittsburgh, Penna. 15213 (U.S.A.) (Received 12 October 1977) (Accepted 14 November 1977)
Summary A statistical reanalysis was performed on the data recently reported on a 6-laboratory, collaborative cytogenetic study to measure and minimize interlaboratory variation. Three of the laboratories had mean values significantly different from the others on most of the 6 indexes of chemically-induced aberration; one laboratory with values higher and two with values lower. Furthermore, relative variability of the values around the means was consistently lower in one of the 6 participating laboratories. The results of the reanalysis of this collaborative study demonstrates that significant interlaboratory differences exist and that these should be adjusted or diminished before rat cytogenetic analysis can be an effective test system for evaluation of a c o m p o u n d for mutagenic potential.
Introduction The results of a "collaborative cytogenetics study to measure and minimize interlaboratory variation" were recently recorded in this journal [1]. 6 laboratories used a joint protocol, c o m m o n glossary and uniform reporting methods. The rats used were from a c o m m o n source. Triethylenemelamine was used at 4 dose levels and results were compared to a control group of untreated animals. In the results of the statistical analyses reported in the article was the statement: "some laboratories were significantly different (P < 0.001) from others." The author of this critique decided to compare the means and measures of variability reported by nonparametric measures most often used for collaborative studies. Abbreviation: CV, coefficient of variation.
286
Statistical methods The nonparametric ranking methods used to intercompare the 6 cytogenetic indexes among the 6 laboratories were the same as those used for the evaluation of interlaboratory consistency of single peroral rat tests [6], rabbit eye and skin irritation tests [ 5] and blood lead analyses [4]. The means or the relative variabilities at each dose level for each index (e.g. chromosome breaks) were ranked, assigning a rank of 1 for the lowest number, 2 for the next l o w e s t . . , and 6 for the highest number reported by any of the 6 laboratories. These ranks were then summed for each laboratory for all of the 4 dose levels for each index. The significance of the sums of ranks was compared to the limits calculated by the methods presented by Thompson and Willke [3] and Youden [7]. As the size of the standard deviation is directly related to that of the mean, a relative measure independent of the size of the unit was used to compare the variability at the 4 dose levels. This measure was the coefficient of variation [2] ; the standard deviation times 100 divided by the mean. Results and discussion The sums of ranks of the means for each index were compared on the raw mean basis as well as for each mean as a change in count from the control value for the respective laboratory. These sums of ranks are presented in Table 1. The data are similar and their significances were similar either for raw data or when these were corrected for control values. Therefore, only the sums of ranks for the raw means, shown in Fig. 1, will be discussed further. The absolute mean and standard deviation values of any of the indexes obtained by the 6 laboratories are presented in the original paper [ 1]. The advantage of ranking procedures is that trends in values for particular laboratories that are higher or lower than the average may be visibly detected. For example, in Fig. 1, laboratory 40 ranked higher (had values higher than the other laboratories) in 5 of the 6 indexes; all but chromosome breaks. With only 6 laboraLo
-~25 ¢~
V 4
/
4
~
4
~
4
4 -Upper 95°1°-8 Limit
°
"G
T'~-"---_.~
c ~: 5 E Chromotid Breaks
-0,2
.
~
~ .
2
~
~
~.3~-L i i Exchange ,,[Proportion [Proportion Figures \ l o f Abnormal ,~,|of Severely ~l Cells \ I Damaged 1 Ceils
o • Lower 95% Limit
~ Aberrations Per Cell
i Chromosome ~reaks
Fig. i. S u m of ranks of m e a n s for 6 indexes of cytogenetic aberrations. N u m b e r s o n graph are laboratory code n u m b e r s f r o m ref. 1; i.e. 0 = laboratory 00, I = laboratory 10, etc.
a Outside l i m i t s ~0.05 = 5 t o 23.
5 3.5 2 1 6 3.5
R a n k of a b o v e s u m of r a n k s f o r e a c h i n d e x 00 5 2 2.5 10 4 1 5 20 1 5 2.5 30 2 4 1 40 6 3 6 80 3 6 4
9 6.5 19 14 12.5 23 a
18 15 10 4 a 22 15
20.5 14 7 10 22 10.5 5 3 2 1 6 4
17.5 13 9.5 7.5 23 a 13.5
__l / Vroport. fProport. I / o f a b n o r - - ~ /~of V m a l cells severely II d a m a g e d ~cells
Aberr. per cell
5 3 1.5 1.5 6 4
19 14 6 6 23 a 16 5 4 1.5 3 6 1.5
20 14 8 12 22 8 2 1 5 4 3 6
9.5 5.5 19 14 13 23 a
Chromosome breaks
2 5.5 3 1 5.5 4
8.5 22 10 7.5 22 14
Exchange figures
5 4 2 1 6 3
18 17 10 4 a 22 13
5 3 2 1 6 4
1S 11 10 8 23 a 14
~/Proport. IVroport. /I ° f a b n ° r " - ~ //° f V m a l cells severely l/damaged ~cells
Chromatid breaks
Exchange figures
Chromatid breaks
Chromosome breaks
S u m of r a n k s of m e a n s c o r r e c t e d f o r c o n t r o l dose a t 4 dose levels
S u m of r a n k s of m e a n s a t 4 d o s e levels
10 20 10 7.5 23 a 13.5
O0 10 20 30 40 80
Laboratory number
RANKS OF MEANS FOR 6 INDEXES OF C Y T O G E N E T I C S A B E R R A T I O N S
TABLE 1
5 3 1.5 1.5 6 4
19 14 6 6 23 a 16
Aberr. per cell
t~ Go
15 14 21 12 14 "8
Chromatid breaks
16 23.5 a 12 12 14.5 6
Chromosome breaks
17 18 12 22 7 8
Exchange figures
15 17 22 17 9 4 a
-~/of abnorce"s
IProport.
P~-oport. of /Ofeverely
15.5 14 16.5 14.5 13 10.5
cells
\ l damaged
S u m of r a n k s of c o e f f i c i e n t s of v a r i a t i o n at 4 d o s e levels
a O u t s i d e limits 0q}.05 = 5 t o 23. b Outside limits ~0.05 = 10 t o 32.
O0 10 20 30 40 80
Laboratory number Aberr.
16 16 16 19 8 9
per ce.
5 3.5 6 2 3.5 1
Chromatid breaks
5 6 2.5 2.5 4 1
Chromosome breaks
4 5 3 6 1 2
Exchange figures
R a n k of s u m of r a n k s f o r e a c h c a t e g o r y
RANKS OF C O E F F I C I E N T S OF V A R I A T I O N FOR 6 I N D E X E S OF C Y T O G E N E T I C A B E R R A T I O N S
TABLE 2
Pr~ort.
Aberr.
3 4.5 6 4.5 2 1
5 3 6 4 2 1
-~ Idamaged ~J cells 4 4 4 6 1 2
of /oa;normoe.s I:fv r01y oer ceB
to 00 00
289 tories and 4 measurements (4 dose levels), a laboratory must have a sum of ranks equal to or less than 5, or 23 or greater, to be statistically significantly different (P < 0.05) for any one index. The maximum possible limits are 4 or 24, therefore, values near the statistical limits may also be indicative of trends. Thus laboratory 40 had sums of ranks of 22, 23, 22, 23 and 23 for chromatid breaks, exchange figures, x/proportion of abnormal cells, x/proportion of severely damaged cells and aberrations per cell, respectively. This result is definitely not a random trend as the sums are all well above the median sum of ranks of 14. Similarly, for these same five indexes, laboratory 20 had sums of ranks of 7, 10, 10, 9.5 and 6 while laboratory 30 had sums of ranks of 10, 7.5, 4, 7.5 and 6, respectively. These sums are also not random and all well below the median sum of ranks of 14. Laboratories 00, 10 and 80 all had sum of ranks for mean values within expected limits for these same 5 indexes. The values for the index of chromosome breaks were plotted, in Fig. 1, separately from the other 5 indexes. Two of the 3 laboratories which had "normal" central tendencies for the other 5 indexes had unexpected low or high trends for chromosome breaks. Laboratory 10 had a sum of ranks of 6.5 for chromosome breaks, as compared to 13, 14, 14, 15 and 20 on the other 5 indexes while laboratory 80 had a sum of ranks of 23, as compared to 10.5, 13.5, 13.5, 15 and 16. The sums of the sums of ranks for the 5 indexes {without that for chromosome breaks) as well as for all 6 measurements are presented in Table 3. Laboratory 40 was statistically higher by any of the comparison procedures while laboratory 30 was significantly lower than expected for the 5 indexes (not including chromosome breaks); laboratory 20 was not statistically significantly different but was always the second lowest of the 6 TABLE 3 SUMS OF SUM OF RANKS AND RANKS OF THESE SUMS Laboratory number
Sum of ranks
Rank of sum of sum of ranks for column
M e a n s at 4 d o s e levels
Means corrected for c o n t r o l d o s e
Coefficients of variation
All 6 indexes
All except chromosome breaks
All 6 indexes
All except chromosome breaks
A
B
C
D
E
24.5 19.5 14 10.5 33 b 24.5
22.5 18.5 9 6.5 a 30 a 18.5
24 20.5 15 11.5 32.5 b 22.5
22 19.5 10 7.5 29.5 a 16.5
26 26 27.5 25 13.5 8 b
A
B
C
D
E
4.5 3 2 1 6 4.5
5 3.5 2 1 6 3.5
5 3 2 1 6 4
5 4 2 1 6 3
4.5 4.5 6 3 2 1
All 6 indexes
Column
O0 10 20 30 40 80
a O u t s i d e l i m i t s ~0.05 = 7 t o 28. b O u t s i d e l i m i t s ~0.05 = 1 0 t o 32.
290 25
~
F | 2
Upper 95% Limit 3
2
o
Chromatid
Breaks
Exchange al Proportion Proportion Figures \1 of Abnormal '~1 of Severely "~ Cells \I Damaged
Aberrations
Chromosome
Per Cell
Breaks
Cells
F i g . 2. S u m o f r a n k s o f c o e f f i c i e n t s o f v a r i a t i o n f o r 6 i n d e x e s of c y t o g e n e t i c a b e r r a t i o n s . N u m b e r s o n ~ a p b are l a b o r a t o r y c o d e n u m b e r s f r o m r e f . 1; i.e. 0 = l a b o r a t o r y 0 0 , 1 = l a b o r a t o r y 10, e t c .
laboratories for the sum of sum of ranks for the means; columns A, B, C and D of Table 3. The relative variabilities were compared by calculation of the coefficients of variation (CV). The lower the coefficient of variation, the lower the variation as a percentage of the mean value for that index. The CVs were ranked and compared among the 6 laboratories exactly as were the means, previously discussed. The ranks for the coefficients of variation are presented in Table 2 and Fig. 2; the sums of sum of ranks are presented in column E of Table 3. Laboratory 80, which had average sum of ranks for the means of all indexes except chromosome breaks, had the lowest or next to lowest CV of the 6 laboratories for each of the 6 indexes of cytogenetic changes; statistically significantly low for the ~/proportion of abnormal cells as well as for all 6 indexes combined (the sum of sum of ranks). Laboratory 40 was the only other laboratory with CVs somewhat different from the others. Laboratory 40 had either the lowest or next to lowest CV for exchange figures, ~/proportion of abnormal cells, ~/proportion of severely damaged cells and aberrations per cell. The relative variations, the CVs, for the other 4 laboratories were approximately the same; the sums of sum of ranks for laboratories 00, 10, 20 and 30 were 26, 26, 27.5 and 25, respectively. The authors of the collaborative study article [1] concluded the following: "this study illustrates that rat cytogenetic analysis can be an effective test system for evaluation of a c o m p o u n d for mutagenic p o t e n t i a l . . . " . This conclusion was reached by them even though they also reported that "some laboratories were significantly different (P < 0.001)." In this current critique, laboratory 40 was consistently higher in mean values for 5 of the 6 indexes examined, while laboratories 20 and 30 had consistently lower mean values; therefore, 3 of the 6 laboratories differed from the others in measurements of the means. Furthermore, laboratory 80 reported relative variability around their mean values lower than those of the other laboratories. Therefore, the reanalysis of the data taken from the collaborative study on cytogenetic variation [ 1 ] demonstrates that statistically significant differences exist between laboratories which may affect the evaluation of the mutagenic potential of a test chemical.
291
This author suggests that the variability between laboratories must be adjusted or diminished before rat cytogenetic analysis can be a reproducible and effective test system for evaluation of a c o m p o u n d for mutagenic potential. References 1 Kilian, D.J., F.M. M o r e l a n d , M.C. Benge, M.S. L e g a t o r a n d E.B. W h o r t o n Jr., A collaborative c y t o genetics s t u d y to m e a s u r e a n d minimize i n t e r l a b o r a t o r y v a r i a t i o n , M u t a t i o n Res., 4 4 ( 1 9 7 7 ) 9 7 - - 1 0 4 . 2 S n e d e c o r , G.W., a n d W.G. C o c h r a n , Statistical Methods, 6 t h edn., t h e I o w a State University Press, Ames, I o w a , 1 9 6 7 , p p . 6 2 - - 6 4 . 3 T h o m p s o n Jr., W.A., a n d T.A. Willke, O n a n e x t r e m e r a n k s u m test f o r outliers, B i o m e t r i k a , 50 ( 1 9 6 3 ) 375--383. 4 Weil, C.S., Critique of i n t e r l a b o r a t o r y e v a l u a t i o n of the reliability of b l o o d - l e a d analyses, A m . Ind. H y g . Assoc. J., 32 ( 1 9 7 1 ) 3 0 4 - - 3 1 2 . 5 Weft, C.S., a n d R . A . Scala, S t u d y o f intra- a n d i n t e r l a b o r a t o r y variability in the results of r a b b i t eye a n d skin i r r i t a t i o n tests, Toxicol. Appl. P h a r m a c o l . , 19 ( 1 9 7 1 ) 2 7 6 - - 3 6 0 . 6 Weft, C.S., a n d G.J. Wright, Intra- a n d i n t e r l a b o r a t o r y c o m p a r a t i v e e v a l u a t i o n o f single oral test, Toxicol. Appl. P h a r m a c o l . , 11 ( 1 9 6 7 ) 3 7 8 - - 3 8 8 . 7 Y o u d e n , W.J., R a n k i n g l a b o r a t o r i e s b y r o u n d - r o b i n tests, Mater. Res. Stand., 3 ( 1 9 6 3 ) 9 - - 1 3 .
Mutation Research, 50 (1978) 285--291 © Elsevier/North-Holland Biomedical Press
A CRITIQUE OF THE COLLABORATIVE CYTOGENETICS STUDY TO MEASURE AND MINIMIZE INTERLABORATORY VARIATION
CARROL
S. W E I L
Carnegie-Mellon Institute of Research, Carnegie-Mellon University, 4400 Fifth Avenue, Pittsburgh, Penna. 15213 (U.S.A.) (Received 12 October 1977) (Accepted 14 November 1977)
Summary A statistical reanalysis was performed on the data recently reported on a 6-laboratory, collaborative cytogenetic study to measure and minimize interlaboratory variation. Three of the laboratories had mean values significantly different from the others on most of the 6 indexes of chemically-induced aberration; one laboratory with values higher and two with values lower. Furthermore, relative variability of the values around the means was consistently lower in one of the 6 participating laboratories. The results of the reanalysis of this collaborative study demonstrates that significant interlaboratory differences exist and that these should be adjusted or diminished before rat cytogenetic analysis can be an effective test system for evaluation of a c o m p o u n d for mutagenic potential.
Introduction The results of a "collaborative cytogenetics study to measure and minimize interlaboratory variation" were recently recorded in this journal [1]. 6 laboratories used a joint protocol, c o m m o n glossary and uniform reporting methods. The rats used were from a c o m m o n source. Triethylenemelamine was used at 4 dose levels and results were compared to a control group of untreated animals. In the results of the statistical analyses reported in the article was the statement: "some laboratories were significantly different (P < 0.001) from others." The author of this critique decided to compare the means and measures of variability reported by nonparametric measures most often used for collaborative studies. Abbreviation: CV, coefficient of variation.
286
Statistical methods The nonparametric ranking methods used to intercompare the 6 cytogenetic indexes among the 6 laboratories were the same as those used for the evaluation of interlaboratory consistency of single peroral rat tests [6], rabbit eye and skin irritation tests [ 5] and blood lead analyses [4]. The means or the relative variabilities at each dose level for each index (e.g. chromosome breaks) were ranked, assigning a rank of 1 for the lowest number, 2 for the next l o w e s t . . , and 6 for the highest number reported by any of the 6 laboratories. These ranks were then summed for each laboratory for all of the 4 dose levels for each index. The significance of the sums of ranks was compared to the limits calculated by the methods presented by Thompson and Willke [3] and Youden [7]. As the size of the standard deviation is directly related to that of the mean, a relative measure independent of the size of the unit was used to compare the variability at the 4 dose levels. This measure was the coefficient of variation [2] ; the standard deviation times 100 divided by the mean. Results and discussion The sums of ranks of the means for each index were compared on the raw mean basis as well as for each mean as a change in count from the control value for the respective laboratory. These sums of ranks are presented in Table 1. The data are similar and their significances were similar either for raw data or when these were corrected for control values. Therefore, only the sums of ranks for the raw means, shown in Fig. 1, will be discussed further. The absolute mean and standard deviation values of any of the indexes obtained by the 6 laboratories are presented in the original paper [ 1]. The advantage of ranking procedures is that trends in values for particular laboratories that are higher or lower than the average may be visibly detected. For example, in Fig. 1, laboratory 40 ranked higher (had values higher than the other laboratories) in 5 of the 6 indexes; all but chromosome breaks. With only 6 laboraLo
-~25 ¢~
V 4
/
4
~
4
~
4
4 -Upper 95°1°-8 Limit
°
"G
T'~-"---_.~
c ~: 5 E Chromotid Breaks
-0,2
.
~
~ .
2
~
~
~.3~-L i i Exchange ,,[Proportion [Proportion Figures \ l o f Abnormal ,~,|of Severely ~l Cells \ I Damaged 1 Ceils
o • Lower 95% Limit
~ Aberrations Per Cell
i Chromosome ~reaks
Fig. i. S u m of ranks of m e a n s for 6 indexes of cytogenetic aberrations. N u m b e r s o n graph are laboratory code n u m b e r s f r o m ref. 1; i.e. 0 = laboratory 00, I = laboratory 10, etc.
a Outside l i m i t s ~0.05 = 5 t o 23.
5 3.5 2 1 6 3.5
R a n k of a b o v e s u m of r a n k s f o r e a c h i n d e x 00 5 2 2.5 10 4 1 5 20 1 5 2.5 30 2 4 1 40 6 3 6 80 3 6 4
9 6.5 19 14 12.5 23 a
18 15 10 4 a 22 15
20.5 14 7 10 22 10.5 5 3 2 1 6 4
17.5 13 9.5 7.5 23 a 13.5
__l / Vroport. fProport. I / o f a b n o r - - ~ /~of V m a l cells severely II d a m a g e d ~cells
Aberr. per cell
5 3 1.5 1.5 6 4
19 14 6 6 23 a 16 5 4 1.5 3 6 1.5
20 14 8 12 22 8 2 1 5 4 3 6
9.5 5.5 19 14 13 23 a
Chromosome breaks
2 5.5 3 1 5.5 4
8.5 22 10 7.5 22 14
Exchange figures
5 4 2 1 6 3
18 17 10 4 a 22 13
5 3 2 1 6 4
1S 11 10 8 23 a 14
~/Proport. IVroport. /I ° f a b n ° r " - ~ //° f V m a l cells severely l/damaged ~cells
Chromatid breaks
Exchange figures
Chromatid breaks
Chromosome breaks
S u m of r a n k s of m e a n s c o r r e c t e d f o r c o n t r o l dose a t 4 dose levels
S u m of r a n k s of m e a n s a t 4 d o s e levels
10 20 10 7.5 23 a 13.5
O0 10 20 30 40 80
Laboratory number
RANKS OF MEANS FOR 6 INDEXES OF C Y T O G E N E T I C S A B E R R A T I O N S
TABLE 1
5 3 1.5 1.5 6 4
19 14 6 6 23 a 16
Aberr. per cell
t~ Go
15 14 21 12 14 "8
Chromatid breaks
16 23.5 a 12 12 14.5 6
Chromosome breaks
17 18 12 22 7 8
Exchange figures
15 17 22 17 9 4 a
-~/of abnorce"s
IProport.
P~-oport. of /Ofeverely
15.5 14 16.5 14.5 13 10.5
cells
\ l damaged
S u m of r a n k s of c o e f f i c i e n t s of v a r i a t i o n at 4 d o s e levels
a O u t s i d e limits 0q}.05 = 5 t o 23. b Outside limits ~0.05 = 10 t o 32.
O0 10 20 30 40 80
Laboratory number Aberr.
16 16 16 19 8 9
per ce.
5 3.5 6 2 3.5 1
Chromatid breaks
5 6 2.5 2.5 4 1
Chromosome breaks
4 5 3 6 1 2
Exchange figures
R a n k of s u m of r a n k s f o r e a c h c a t e g o r y
RANKS OF C O E F F I C I E N T S OF V A R I A T I O N FOR 6 I N D E X E S OF C Y T O G E N E T I C A B E R R A T I O N S
TABLE 2
Pr~ort.
Aberr.
3 4.5 6 4.5 2 1
5 3 6 4 2 1
-~ Idamaged ~J cells 4 4 4 6 1 2
of /oa;normoe.s I:fv r01y oer ceB
to 00 00
289 tories and 4 measurements (4 dose levels), a laboratory must have a sum of ranks equal to or less than 5, or 23 or greater, to be statistically significantly different (P < 0.05) for any one index. The maximum possible limits are 4 or 24, therefore, values near the statistical limits may also be indicative of trends. Thus laboratory 40 had sums of ranks of 22, 23, 22, 23 and 23 for chromatid breaks, exchange figures, x/proportion of abnormal cells, x/proportion of severely damaged cells and aberrations per cell, respectively. This result is definitely not a random trend as the sums are all well above the median sum of ranks of 14. Similarly, for these same five indexes, laboratory 20 had sums of ranks of 7, 10, 10, 9.5 and 6 while laboratory 30 had sums of ranks of 10, 7.5, 4, 7.5 and 6, respectively. These sums are also not random and all well below the median sum of ranks of 14. Laboratories 00, 10 and 80 all had sum of ranks for mean values within expected limits for these same 5 indexes. The values for the index of chromosome breaks were plotted, in Fig. 1, separately from the other 5 indexes. Two of the 3 laboratories which had "normal" central tendencies for the other 5 indexes had unexpected low or high trends for chromosome breaks. Laboratory 10 had a sum of ranks of 6.5 for chromosome breaks, as compared to 13, 14, 14, 15 and 20 on the other 5 indexes while laboratory 80 had a sum of ranks of 23, as compared to 10.5, 13.5, 13.5, 15 and 16. The sums of the sums of ranks for the 5 indexes {without that for chromosome breaks) as well as for all 6 measurements are presented in Table 3. Laboratory 40 was statistically higher by any of the comparison procedures while laboratory 30 was significantly lower than expected for the 5 indexes (not including chromosome breaks); laboratory 20 was not statistically significantly different but was always the second lowest of the 6 TABLE 3 SUMS OF SUM OF RANKS AND RANKS OF THESE SUMS Laboratory number
Sum of ranks
Rank of sum of sum of ranks for column
M e a n s at 4 d o s e levels
Means corrected for c o n t r o l d o s e
Coefficients of variation
All 6 indexes
All except chromosome breaks
All 6 indexes
All except chromosome breaks
A
B
C
D
E
24.5 19.5 14 10.5 33 b 24.5
22.5 18.5 9 6.5 a 30 a 18.5
24 20.5 15 11.5 32.5 b 22.5
22 19.5 10 7.5 29.5 a 16.5
26 26 27.5 25 13.5 8 b
A
B
C
D
E
4.5 3 2 1 6 4.5
5 3.5 2 1 6 3.5
5 3 2 1 6 4
5 4 2 1 6 3
4.5 4.5 6 3 2 1
All 6 indexes
Column
O0 10 20 30 40 80
a O u t s i d e l i m i t s ~0.05 = 7 t o 28. b O u t s i d e l i m i t s ~0.05 = 1 0 t o 32.
290 25
~
F | 2
Upper 95% Limit 3
2
o
Chromatid
Breaks
Exchange al Proportion Proportion Figures \1 of Abnormal '~1 of Severely "~ Cells \I Damaged
Aberrations
Chromosome
Per Cell
Breaks
Cells
F i g . 2. S u m o f r a n k s o f c o e f f i c i e n t s o f v a r i a t i o n f o r 6 i n d e x e s of c y t o g e n e t i c a b e r r a t i o n s . N u m b e r s o n ~ a p b are l a b o r a t o r y c o d e n u m b e r s f r o m r e f . 1; i.e. 0 = l a b o r a t o r y 0 0 , 1 = l a b o r a t o r y 10, e t c .
laboratories for the sum of sum of ranks for the means; columns A, B, C and D of Table 3. The relative variabilities were compared by calculation of the coefficients of variation (CV). The lower the coefficient of variation, the lower the variation as a percentage of the mean value for that index. The CVs were ranked and compared among the 6 laboratories exactly as were the means, previously discussed. The ranks for the coefficients of variation are presented in Table 2 and Fig. 2; the sums of sum of ranks are presented in column E of Table 3. Laboratory 80, which had average sum of ranks for the means of all indexes except chromosome breaks, had the lowest or next to lowest CV of the 6 laboratories for each of the 6 indexes of cytogenetic changes; statistically significantly low for the ~/proportion of abnormal cells as well as for all 6 indexes combined (the sum of sum of ranks). Laboratory 40 was the only other laboratory with CVs somewhat different from the others. Laboratory 40 had either the lowest or next to lowest CV for exchange figures, ~/proportion of abnormal cells, ~/proportion of severely damaged cells and aberrations per cell. The relative variations, the CVs, for the other 4 laboratories were approximately the same; the sums of sum of ranks for laboratories 00, 10, 20 and 30 were 26, 26, 27.5 and 25, respectively. The authors of the collaborative study article [1] concluded the following: "this study illustrates that rat cytogenetic analysis can be an effective test system for evaluation of a c o m p o u n d for mutagenic p o t e n t i a l . . . " . This conclusion was reached by them even though they also reported that "some laboratories were significantly different (P < 0.001)." In this current critique, laboratory 40 was consistently higher in mean values for 5 of the 6 indexes examined, while laboratories 20 and 30 had consistently lower mean values; therefore, 3 of the 6 laboratories differed from the others in measurements of the means. Furthermore, laboratory 80 reported relative variability around their mean values lower than those of the other laboratories. Therefore, the reanalysis of the data taken from the collaborative study on cytogenetic variation [ 1 ] demonstrates that statistically significant differences exist between laboratories which may affect the evaluation of the mutagenic potential of a test chemical.
291
This author suggests that the variability between laboratories must be adjusted or diminished before rat cytogenetic analysis can be a reproducible and effective test system for evaluation of a c o m p o u n d for mutagenic potential. References 1 Kilian, D.J., F.M. M o r e l a n d , M.C. Benge, M.S. L e g a t o r a n d E.B. W h o r t o n Jr., A collaborative c y t o genetics s t u d y to m e a s u r e a n d minimize i n t e r l a b o r a t o r y v a r i a t i o n , M u t a t i o n Res., 4 4 ( 1 9 7 7 ) 9 7 - - 1 0 4 . 2 S n e d e c o r , G.W., a n d W.G. C o c h r a n , Statistical Methods, 6 t h edn., t h e I o w a State University Press, Ames, I o w a , 1 9 6 7 , p p . 6 2 - - 6 4 . 3 T h o m p s o n Jr., W.A., a n d T.A. Willke, O n a n e x t r e m e r a n k s u m test f o r outliers, B i o m e t r i k a , 50 ( 1 9 6 3 ) 375--383. 4 Weil, C.S., Critique of i n t e r l a b o r a t o r y e v a l u a t i o n of the reliability of b l o o d - l e a d analyses, A m . Ind. H y g . Assoc. J., 32 ( 1 9 7 1 ) 3 0 4 - - 3 1 2 . 5 Weft, C.S., a n d R . A . Scala, S t u d y o f intra- a n d i n t e r l a b o r a t o r y variability in the results of r a b b i t eye a n d skin i r r i t a t i o n tests, Toxicol. Appl. P h a r m a c o l . , 19 ( 1 9 7 1 ) 2 7 6 - - 3 6 0 . 6 Weft, C.S., a n d G.J. Wright, Intra- a n d i n t e r l a b o r a t o r y c o m p a r a t i v e e v a l u a t i o n o f single oral test, Toxicol. Appl. P h a r m a c o l . , 11 ( 1 9 6 7 ) 3 7 8 - - 3 8 8 . 7 Y o u d e n , W.J., R a n k i n g l a b o r a t o r i e s b y r o u n d - r o b i n tests, Mater. Res. Stand., 3 ( 1 9 6 3 ) 9 - - 1 3 .
