t
30
0
/
I 1 I
I '
I
I I I I
I I
1
7'
"1
2
10
/
I
/
AI
I
/
//
/A
5 mg CI - ( 0
1,
mg CI ( A
I
A/
/
d'
i,
that tobacco smoke may also contain other unsaturated, chlorinated hydrocarbons with carcinogenic activity. This may include the suspected human carcinogen 2-chlorobutadiene (chloroprene; 11, 12) as well as chlorinated products of isoprene, the major unsatured hydrocarbon in tobacco smoke. NOTE ADDED IN PROOF.This study was presented a t the 29th Tobacco Chemists' Research Conference (College Park, Md., Oct. 8-10, 1975). Two questions arose. 1) How much VC is in ambient air? Using the above method, a t different times we passed 20 1. of ambient laboratory air through charcoal. The extract of the charcoal was each time free of VC. 2) Is VC artificially formed from the ethylene of the smoke (trapped on charcoal) and the trace amounts of chlorine possibly present in the bromine? To answer this question, we led 400 ml of ethylene (CP-Matheson Gas Products) through charcoal and, using the same bromine, we processed it in the usual way. We did not observe a signal in the glc a t the retention time of DBVC using the ECD a t twice the usual sensitivity.
9
1 I
10 per
g
15
D r y Tobacco
Figure 3. Correlation between inorganic chloride and total chlorine in tobacco and vinyl chloride in cigarette smoke
ACKNOWLEDGMENT The authors thank C. Costello of the Mass Spectroscopy Laboratory, Massachusetts Institute of Technology, for her cooperation in obtaining and interpreting mass spectral data on DB-VC. LITERATURE CITED
tion of inorganic chloride in tobacco ( 1 0 ) . Figure 3 indicates that water soluble chloride, as well as total chlorine, in tobacco may contribute to the concentration of VC in the smoke. We realize that detailed studies are needed to ensure the nature of the major precursors in tobacco for the chlorine of VC in the smoke. Nevertheless, we should not disregard the possibility that also during the burning of other plant materials, and in fact of all organic matter which contains chlorine, vinyl chloride can be generated and can be released into our respiratory environment, as demonstrated for marijuana smoke. This study has shown that smoke contains up to 16 ng of vinyl chloride per cigarette, and 27 ng of VC per little cigar. This amount corresponds to a concentration of about 30 ppb. Based on human data and results from animal studies, it appears to us that these minute amounts of VC will not contribute to a measurable degree to the carcinogenic activity of tobacco smoke. I t should be realized, however, that we have directed our analysis only toward the identification and quantitative determination of vinyl chloride and
(1) International Agency for Research on Cancer, Monogr. 7,291 (1974). (2) I. J. Sellkoff and E. C. Hammond. "Toxicity of Vinyl Chloride-Polyvinyl Chloride", Ann. N.Y. Acad. Sci., 248, 1 (1975). (3)C. Maltoni, G. Lefemine, P. Checo, and D. Carretti, Ospedalidi, Bologna, ltay, I (5-6),l(1974). (4) J. R. Caldwell and H. V. Moyer, Anal. Chem., 7,38 (1935). (5) M. Yamazaki and J . Yamazaki, Sci. Pap. Central Res. lnst. Jpn Monopolycorp., 114, 45 (1972). (6)E. L. Wynder and D. Hoffmann. "Tobacco and Tobacco Smoke, Studies in Experimental Carcinogenesis", Academic Press, New York, N.Y., 1967. (7)F. F. Guthrie and P. G. Bowery, Residue Rev., 10, 31 (1967). (8)R. L. Stedman, Chem. Rev,, 88, 153 (1968). (9)T. C. Tso. "Physiology and Biochemistry of Tobacco Plants", Dowden, Hutchinson, and Ross, Stroudsburg, Pa., 1972. (10)R. R. Johnson and T. E. Smith, Abstr. 24th Tobacco Chemists' Res. Conf., 24 (1970). ( 1 1 ) E. A. Khackatryan, Gig. Tr. Prof. Zaboi., 18, 54 (1972). (12) E. A. Khachatryan, Pub\. Oncol. USSR., 18, 85 (1972)
RECEIVEDfor review August 5, 1975. Accepted September 22, 1975. This is No. XXXVI of "Chemical Studies on Tobacco Smoke". I t was supported in part by American Cancer Society Grant BC-56T and by Public Health Service Contract ECI-SHP-74-106.
Thin Layer Chromatography/Densitometry with Transferable Calibration Factors H. Bethke and R. W. Frei" Analytical Research a n d Development, Pharmaceutical Division, Sandoz Ltd., 4002 Basle, Switzerland
The principle of a transferable calibration technique is discussed and introduced with examples of dihydroergotamine. The quantitation of spots is carried out wlth direct UV refiectance spectrometric measurements of methylergobasine on different serles of chromatopiates. The appiicabiiity of this same calibration transfer model to in situ fiuorescence measurements which require a derivatizating reac50
ANALYTICAL CHEMISTRY, VOL. 48, NO. 1, JANUARY 1976
tion after separation is demonstrated with samples of digoxin. The reproducibility of the results obtained Is comparable to indlviduai calibration carried out for each plate (rei std dev about 2 % ). With the use of transferable calibration, 5 analyses can be carried out on one 20 X 20 cm plate which represents an increase in capacity of 60% compared to a previously discussed method.
The use of a data pair technique ( I ) enables the chromatographer to compensate for fluctuating chromatographic conditions such as varying layer thickness or edge effects. This internal compensation approach works by pairing up the measurements of two spots on the same plate; one from the edge and the other from the center. A plate with n spots yields n / 2 data pairs. The resulting improvement in reproducibility is highly significant and % re1 std dev values for complete analysis of pharmaceutical preparations between 1.1 and 1.9 were reported ( 2 ) . TLC with identification by UV reflectance spectrometry was thus recommended as a truly quantitative technique for routine analysis of pharmaceuticals. In this method and in most other densitometric TLC techniques (31, each plate has to carry its proper set of standards for quantitative measurements. Klaus ( 4 , 5 ) has described a quantitative method with transferable calibration curves by means of an internal standard, but generally the fluctuations from one plate to another are too large for the use of one common calibration curve for a series of different chromatoplates. With the application of 1 2 spots according t o the datapair method, there is room for three analyses as a result of double application and calibration with three reference solutions. If the number of reference spots for each calibration on the plate can be reduced, then the analytical capacity can be increased to the same extent. Our efforts were therefore directed toward achieving a maximum of external transferable information and a minimum of individual calibration-namely, only one calibration point on each TLC plate.
EXPERIMENTAL All chromatographic developments have been carried out on commercially coated silica gel Merck F 2 j q plates (Merck, Darmstadt, G.F.R.). Spot application was made with Drummond disposable Microcaps (Drummond, Broomall, Pa.), 2 or 5 11. All solvents were of reagent grade. The following developing systems were used: for methylergobasine: chloroform-ethanol-ammonia (70:30:2); for dihydroergotamine: dichloromethane-methanolammonia (80:15:0.1); for digoxin: ethylacetate-dichloromethaneisopropanol-ammonia (100:60:40:13). The plates were dried a t room temperature. No significant changes in measured reflectance have been noticed with the systems investigated after 1-hr exposure of the plates to UV light. The fluorescence of derivatized digoxin was stable within the time of measurement (about 30 rnin). For the derivatization of digoxin, methods described by W. Messerschmidt ( 6 ) and J. Frijns (7) were adapted and the reaction was carried out according to the following procedure. After chromatographic development the plates were heated up to 180 OC for 10 min and then exposed to hydrochloric acid vapor (5 min in a chromatographic chamber containing concentrated hydrochloric acid). After the reaction, the plates were again exposed to 180 "C (90 rnin). Excess HC1 was removed by evacuation of the drying oven. Before measurement, the plates were cooled to room temperature for 1 or 2 hr. Densitometric measurements were carried out with a Zeiss chromatogram spectrophotometer (Zeiss, Oberkochen, Wiirttemberg, G.F.R.), an Infotronics Model CRS 104 integrator (Infotronics, Boulder, Colo.) and a W & W Model 2211 recorder (W & W Electronic, Miinchenstein, Switzerland). The densitometer was set to M-Pr arrangement. Methylergobasine was measured a t 320 nm, dihydroergotamine at 280 nm. In situ fluorimetric measurements were carried out with a Perkin-Elmer MPF3 fluorimeter with TLC-accessory (Hitachi-Perkin-Elmer) and a Perkin-Elmer S I P integrator (Perkin-ElmerBodenseewerk, Ueberlingen, G.F.R.). The settings of the fluorimeter were: ,A, 345 nm, A,, 440 nm, excitation slit 20 nm, emission slit 25 nm.
DISCUSSION OF THE PRINCIPLE The difficulties of external or transferable calibration in TLC are well known. The slopes of calibration curves (A = a + b c ) obtained for a given range of concentrations (with-
Calibration plots A = a methylergobasine Figure 1.
+ bc for four different series of
in which the expected value for the sample being analyzed lies) vary greatly from plate to plate and the curves do not usually pass through the origin of coordinates (Figure 1). The result is that complete external calibration is impossible and that it is necessary to have a t least two measured points to determine the coefficients for slope ( b ) and the ordinate intercept ( a ) . Under the most ideal conditions imaginable-either slopes ( b ) being parallel or ordinate intercepts ( a ) the same, within the normal limits of error-the procedure would be simple and the appropriate coefficients could be taken over directly. However, as can be seen in Figure l e for the analysis of methylergobasine by UV reflectance, the variations of a and b are too great. If, however, one looks more closely (Figure 1, a-d), one sees that the tangle of calibration curves clears somewhat. The calibration curves of plates which were prepared and measured in separate series a t different times, were grouped together and it was found that while they differ greatly from series to series, they lie closer together within the same series. At the same time, there is a tendency for the curves in each series to vary less in slopes b than in intercept a . On the other hand, the mean slope ( b ) of each group can vary from series to series (Table I). In the light of these results, we tried to use the slope h for the external calibration. Two approaches were used for this transfer. Direct Transfer. Since, a t least within a given series and to some extent also between different series, the slope coefficients lie close together, the mean slope b was selected as the direct transfer factor and the ordinate intercept a was then calculated on the individual plate wit,h the single calibration value ( A x ,c.J. A schematic presentation for this procedure is given in Figure 2. Transfer of the Relative Slope. In order to compensate better for the jump in the mean slope value between series, value b for each plate was divided by a reference measured value ( A i d ) , ideally by the mean value of the calibration reANALYTICAL CHEMISTRY, VOL. 48, NO. 1, JANUARY 1976
* 51
Table I. Comparison o f S e a n Slope Values and Mean Relative Slope Values brel for Different Series of Methylergobasine Series
4
brei 0.232 0.266 brei = 0.238 0.227 s = 17.7% 0.229 0.209 0.170 brei = 0.204 0.235 s = 113.1% 0.201 0.262 0.248 0.225 brei = 0.234 0.239 s = 110.6% 0.243
b
1.74 2.13 1.76 1.74 2.54 1.99 3.17 2.99 4.35 3.98 3.17 3.47 3.13 2.75 3.66 3.38 4.02
1
b = 1.84 s = 110.4%
6 = 2.67
s = ?19.7%
6 = 3.47
s = 117.1%
r e l a t i v e slope
i;;;;
ASt3 Ast2 Astl
i
0.190
0.301 0.294
6 = 3.44
0.265 o.208 0.273
3*68 s = 111.6%
3.45 3.13 2.79 1-4
Transfer of th.
=
0.252
s = 116.1% indlvldual plate
0.216 0.209
:
0 ” ~Callbration value Ax
6 = 3.00 s = 126.0%
=
0.235
’
cx
transfered
.I
s = t 14.1%
g&. These relatiue values (brei)were then again averaged ( b r e J and used as the transfer factor. Then for computing the coefficients of the individual calibration curves on each plate, slope b was first calculated from the measured valueof the single calibration spot by back multiplication with brei and then ordinate distance a in the same way as described for the direct transfer. Figure 3 gives the schematic presentation of this procedure.
meaiueed
known
I
i
computed
Figure 3. Schematic presentation of the principle of transferred calibration A, = peak area of standards 1. 2, 3. A,d = mean areafthe calibration region. b,,, = relative slope for each plate i,2 . . , n. 4.1 = mean relative slope value. b, = individualslope. a, = individualordinate intercept
Direct transfer of ;d
several plates
:
:
RESULTS
= al + bl
A
=a2+bZ.C
A
zan+bn.C
C
E b.
-E= individual plate
.
A
I n
one calibration AX i
mefsured
value
cx
transfered kTown
t
computed
Figure 2. Schematic presentation of the principle for direct transfer A = peak area, b = slope, b = mean slope value, a = ordinate intercept, a, = individual ordinate intercept, c = concentration
52
ANALYTICAL CHEMISTRY, VOL. 48, NO. 1, JANUARY 1976
If one compares the variations between the two transfer factors in the above examples (Table I), it can be seen that transfer of the relative slope gives greater uniformity in most individual series, but even more for a whole complex of several series measured at different times. T o discuss the practical consequences of this method, a single series in the analysis of a dihydroergotamine ampoule solution was investigated (Figure 4). To exclude sources of error due to chromatography or measuring techniques, normal analytical plates were used and evaluated only by the methods described. As can be seen in Figure 4, the individual calibration curves, with one exception, are really close. Table I1 gives the correlation coefficients of the five curves (determined from three calibration values) and the relative analytical errors resulting from the transferred calibration in comparison to the direct individual calibration on each plate; they are given as the means of three analyses per plate. If one compares t h g i r s t two of the columns for hU,,i (calibration with 6 and brei), it is clear that there is no significant difference between direct transfer of b and calibration with the relative value brei. This result agrees with the previous finding that the gain resulting from the formation of a relative value within a series is small.
Table 11. Correlation Coefficients and Relative Analytical Errors A U r e l Obtained by Comparing Transferred Calibration to Direct Calibration (Mean Values for Three Analyses of Dihydroergotamine Drop Solution per Plate) -
Calibra- Correlation tion coefficient, plot Y
1 2 3 4 5
AUrela calibration w i t h
6, % b
0.9976 0.9991 0.9906 0.9995 0.9996
c o n c e n t r a t i o n reg i o n used f o r calibration
Kl,7ob b"), %c
2.8 1.0 2.7 -0.8 0.4
1.3 1.0 3.9 -0.8 0.8
b re1 ( 5 )1 %'
1.1 1.1 3.4 -0.8 0.7
3.7 0.7 3.5 -0.6 0.7
a A u r e l = (Utransferred calibration ' ' ' calibration)/ - - - U individual Ulndivldual , . . calibration x 100. b b ; brei = mean values of all 5 plates. c b ( 5 ) ;b t ) = values only from plate 5. I
Dihydroergotamine
Table 111. Mean Values of the Relative Errors of 37 Analyses of Methylergobasine Ampoules
1 Aure1 1%
Figure 4. Calibration plots for the analysis of dihydroergotamine ampoule solutions
values after transferred calibration
Transferred value
4 separated groups
Total series
G
0.9% 0.6%
1.2% 0.6%
b
a l A c r r e l l = jutransf. calibr.
- Uindiv. calibr.l/Uindiv. calibr.
z.
X
100.
Table IV. Comparison of Results Obtained for Methylergobasine Ampoules by the Transferred Calibration and Individual Calibration Techniques Reproducibility/methylergobasine Li,
Plate
1
2 3 4 1-4
[%I
3
transferred calibration
G2,[,%I1
individual calibration
103.2 103.6 101.4 101.7 103.3 103.6 104.9 105.9 103.2 104.5 104.1 103.6 105.2 104.8 105.2 104.8 108.3 107.3 107.1 106.1 107.7 106.7 U ,= 1 0 4 . 9 % U , = 104.8% = t2.0% s = t 1.6% - c', = 0.1%
s
of curve 5 were chosen as transfer factors rather than the means b or Curve 5 has good correlation, so no secondary error is incorporated, but its slope b is the least steep. As expected, the results obtained from this variation using the relative slope value are affected less than those using direct transfer. On the basis of these results, one can now consider the treatment of a larger series such as depicted in Figure L. This result is shown in Table I11 which gives a comparison between the mean values of the relative errors of 37 analyses after transferred calibration, once from the point of view of direct and relative transfer and once after differentiating between the four individual series and the total of all four series. For the transfered relative calibration, there is no difference between separated groups and the total series and the result is satisfactory. With the direct transfer, however, the mean error is twice as high. To demonstrate the comparability obtained by the use of transferred mean relative slopes even more clearly, another example is given of the reproducibility obtained with a series of methylergobasine analyses (Table IV) by comparing only individual calibration and transferred relative calibration. In a series with 4 plates and 11 analyses of the same batch, practically identical means were found and the standard deviations did not differ from one another ( F test: p = 95%). The differences between transferred and individual calibration for each sample are not significant either. In a continuation of this study, it was also attempted to test the applicability of the transferred calibration approach to systems which require derivatization after chromatographic separation and to systems which are evaluated by fluorescence densitometry rather than by UV measurements. The analysis of digoxin by direct fluorescence evaluation after derivatization was chosen for this purpose. In Figure 5 one can see the calibration curves of the series of measurements; again there is quite a wide scatter. The slope values ( b ) and the relative slopes (brei) are shown in Table V. The marked improvement for the uniformity of mean slopes obtained by relativization is clear. In Table VI, the comparison between the two calibration methods is shown by means of the mean relative error of 36 analyses of digoxin ampoule solutions. In this example too,
u,- u, -0.33 -0.30 -0.33 -1.03 -1.27 0.56 0.36 0.36 1.00 0.98 0.98 U , - U , = 0.68
- u,
In addition, it is clear that for some plates the deviations from individual calibration are insignificant, although of some importance in two plates (1 and 3). For plate 3, both results are poor. This plate had the poorest correlation coefficient for the calibration curve and, on examination of all the individual figures, it was observed that just the calibration value used for the back calculation deviated markedly from the norm. In the case of curve 1, only direct transfer produces unsatisfactory results. Figure 4 shows that curve 1 has the steepest slope bl and the deviation from the mean transfer value is therefore greatest for direct transfer. It is precisely this problem that is increased in the treatment of a number of series and which already can be demonstrated in the last two columns of Table 11, for which only the extreme values
ANALYTICAL CHEMISTRY, VOL. 48, NO. 1, JANUARY 1976
53
Table V. Comparison of Mean Slope Values 6 and Mean Relative Slope Values brei for Different Series of Digoxin Series
b
1
4.99 6.07
brei
- = 4.84
3,53 b 4.76 2
9.24
0.793
11'78
12.71 15.51 1 4 ..~ 30
3
= 12.36
0800
= 13.11
0.762 0.825 0.7 6 3 0.749
~
9.47 12.64 16.05
0.809 0.604 - = 0 , 7 2 4 brel 0.771
-
0:898 0.847
b = 10.10 s = i43.8%
1-3
0.727
brei = 0.835 -= 0 . 7 7 5 brei
-
brei = 0 . 7 7 9 s = T9.3%
Table VI. Mean Values of the Relative Errors of 36 Analyses of Digoxin Ampoules lAUrel Transferred value
-
1'-
value after transferred calibration 3 separated groups
brei ' I A v r e l 1 = 1 utransf. calibr. -uindiv. b
Total series
1.8% 0.7%
2.9% 0.7% calibr.1 luiiindiv. calibr.
100. Figure 5.
Calibration plots of three different series for digoxin
no significant difference is discernible for the serial transfer or the total transfer-as long as the relative value is taken! Moreover, these errors are all tolerable and lie within the limits normally present in densitometry, even with good individual calibration.
given as a condition, we are, however, covering all routine applications in pharmaceutical production control. Five analyses can be performed on one thin layer plate using the data-pair principle, which represents an increase in capacity of 60%.
CONCLUSION
ACKNOWLEDGMENT
The experimental results demonstrate that the system of transferred calibration with the use of relative slope values can be used to achieve reproducible results comparable with individual calibration in direct thin layer evaluation. Considering the example with the poor correlation coefficient (Table 11),one can see that the prerequisite is that, in a nondeviant test series, the transfer factor must be determined while parallel multiple analyses guarantee that any outliers are revealed. An absolute condition for the general applicability is that the transfer factors be determined in the same concentration range in which the unknown concentration is to be expected (f20%).This means that the validity of this method becomes questionable with a series of sample concentrations which are scattered over a wide concentration range. With the tolerance limit of f 2 0 %
54
ANALYTICAL CHEMISTRY, VOL. 48, NO. 1, JANUARY 1976
C. Karsenty, M. Moustache, and D. Redey are acknowledged for carrying out the experiments. LITERATURE CITED (1) H. Bethke and R. W. Frei, J. Chromatogr. Sci., 12, 392 (1974). (2) H. Bethke and R. W. Frei, J. Chromatogr., 91, 433 (1974). (3) R. W. Frei, "Reflectance Spectroscopy in TLC" in "Progress in TLC and Related Methods", A. Niederwieser and G. Pataki, Ed., Vol. 11, Ann Arbor Science Publishers, Ann Arbor, Mich., 1971. (4) R. Klaus, J. Chromatogr., 40, 235 (1969). (5) R. Klaus, J. Chromatogr., 62, 99 (1971). (6) W. Messerschmidt, Dtsch. Apoth. Ztg, 110, 359 (1970). (7) J. Frijns, Pharm. Weekblad, 105, 209 (1972).
RECEIVEDfor review April 4, 1975. Accepted September 29, 1975. Presented a t the 122nd Annual Meeting of the American Pharmaceutical Association, San Francisco, Calif., April 1975.
