- Email: [email protected]
Expired air carbon monoxide and saliva thiocyanate: Relationships to self-reports of marijuana and cigarette smoking
Addictive Behaviors, Vol. 10, pp. 137-144, 1985 Printed in the USA. All rights reserved.
0306-4603185 $3.00 + .OO Copyright e 1985 Pergamon Press Ltd
EXPIRED AIR CARBON MONOXIDE AND SALIVA THIOCYANATE: RELATIONSHIPS TO SELF-REPORTS OF MARIJUANA AND CIGARETTE SMOKING ANTHONY
BIGLAN, CHERI GALLISON, ROBERT THOMPSON
DENNIS ARY, and
Oregon Research Institute Abstract-This study examined the relationship between self-reports of marijuana and cigarette smoking, and the physiological measures of expired air carbon monoxide (CO) and saliva thiocyanate (SCN) in a sample of 1,130 seventh, ninth, tenth, eleventh, and twelfth graders. Subjects who reported marijuana smoking were likely to also report cigarette smoking. The correlation between self-reported marijuana smoking and SCN was negligible. The correlation between CO and self-reported marijuana smoking was statistically significant, but when the variance due to cigarette smoking was removed, this relationship also became negligible. The existence of a sizeable number of marijuana smokers in this sample of adolescents did not alter the correlation between CO and self-reports of cigarette smoking. However, in adult samples, where marijuana and cigarette smoking are less highly correlated, marijuana smoking could affect the relationship between CO and self-reported cigarette smoking. Carbon monoxide predicted self-reported cigarette smoking better than did saliva thiocyanate. There was an interaction between grade and the CO/cigarette smoking correlation. The correlations were generally higher in upper grades.
This article examines the relationships between two physiological measures and adolescents’ self-reports of marijuana and cigarette smoking in a sample of 7th, 9th, lOth, llth, and 12th graders. Expired air carbon monoxide (CO) and saliva thiocyanate (SCN) have been shown to correlate with self-reports of adolescent cigarette smoking (Pechacek, Murray, Luepker, Mittelmark, Johnson, & Shultz, 1984). However, the possible impact of marijuana on these measures has not been examined among subjects who routinely smoke marijuana. The CO and SCN levels associated with self-reported marijuana smoking are of interest for two reasons. First, if there are strong correlations between reports of marijuana smoking and physiological measures, the measures could be useful in corroborating reports of marijuana smoking, especially among persons who do not smoke cigarettes. Second, because marijuana smoking appears to occur among a significant proportion of young people (Penning & Barnes, 1982), it may affect the validity of these physiological measures as assessments of adolescent cigarette smoking. Thus, in the present study the relationship between self-reported marijuana and cigarette smoking is examined and the effect of marijuana smoking on the correlations between cigarette smoking and CO and SCN are assessed. Pechacek et al. (1984) have presented evidence suggesting that the relationships between SCN and CO and the number of cigarettes smoked is greater with increasing age. This implies that younger smokers may receive a smaller dose per cigarette and that Portions of this paper were presented at the American Psychological Association, Anaheim, 1983. This study was supported by Grant #HD 13409 and Grant #HD 15825 from the National Institute of Child Health and Human Development. The authors wish to thank Agatha McLean for typing assistance and Dayna Mitchell, LaVonne Boss, and Kathy Fisher for assistance in data collection. Requests for reprints should be sent to Anthony Biglan, Oregon Research Institute, 195 West 12th Ave., Eugene, OR 97401. 137
138
ANTHONY
BIGLAN
et al.
adolescents may take longer to develop adult-type inhalation patterns than was previously assumed. In the present study, the replicability of the Pechacek et al. (1984) finding is assessed and the following rival explanations are examined empirically. First, the results may have been due to the smaller proportion of smokers in lower grades and the fact that smoking rates among younger smokers were more restricted. Both of these factors would tend to reduce the correlation between self-reported cigarette smoking and CO or SCN. Second, Pechacek et al. (1984) identified several subjects in the lower age groups who may have been exaggerating how much they smoked. This would produce lower correlations.
METHOD
Subjects Subjects were 1,130 students who were participating in an experimental evaluation of a smoking prevention project. They were distributed across grades as follows: 455 seventh graders, 346 ninth graders, 150 tenth graders, and 179 eleventh and twelfth graders. The complete sample included 573 females and 557 males. The majority of the subjects analyzed (87%) were being assessed at 6-month follow-up with (13%) assessed prior to any intervention. The sample included three high schools and four middle schools. Measures Expired air carbon monoxide. The expired air sample was assessed for CO content using an Energetic Science Ecolyzer 2000. The assessment occurred within 12 hours of sample collection. Saliva thiocyanate (SCN). Saliva samples were collected in a sterile Falcon 17 x 100, 16 mm test tube using 1%” sterile dental cotton rolls. Saliva samples were stored within eight hours of collection in a professional freezer locker at - .18”F. For analysis, they were shipped frozen to the University of Minnesota Health Behaviors Laboratory where saliva thiocyanate levels were determined by spectophotometer (Luepker, Pechacek, Murry, Johnson, Hurd, & Jacobs, 1981). Because of the expense of analyzing saliva thiocyanate samples, it was necessary to sample subjects on this variable. All self-reported smokers of cigarettes and/or marijuana were selected for analysis from the total sample along with 25% of the self-reported nonsmoking subjects. The samples were stored for 2 to 8 months before shipping. Questionnaire. Current self-reported smoking was assessed by asking subjects to write down the number of cigarettes they smoked in the last week and the number of cigarettes they smoked in the previous 24 hours. A weighted average of these items(Cigarettes last week + 7 x cigarettes yesterday)/2 - was composed. Marijuana smoking was assessed by asking subjects to indicate the number of times they smoked marijuana in the last week. Subjects could choose a number 0 through 10 or the item “more than ten times.” For the purposes of analysis, the latter response was coded as eleven. Procedures Two weeks prior to data collection, a researcher described the project in each classroom that was to be assessed. Students were told that the assessment consisted of completing a questionnaire regarding their smoking experiences, and supplying breath and saliva samples. Simultaneously, a letter was mailed to the parents, explaining the as-
Measurement of marijuana and cigarette smoking
139
sessment for which their child was volunteering. Parents who did not want their child to participate returned a postcard indicating their declination. Two or three staff members of the smoking prevention project collected the data in the classroom setting. Subjects began by filling out an identification sheet. Following this, they provided an expired air carbon monoxide (CO) sample. Prior to collection of the breath sample, subjects were informed that analysis of expired air for carbon monoxide (CO) content would allow identification of cigarette smokers. This was done because there is evidence that such instructions increase the accuracy of self-reports of cigarette smoking (Evans, Hansen, & Mittlemark, 1977; Bauman & Dent, 1982). The air sample collection procedure involved subjects holding their breath for 15 seconds without taking in additional air. After 15 seconds, subjects breathed into a l- or 2-liter polyethylene bag, then securely sealed it with a rubberband. Next, subjects provided a saliva sample using a procedure similar to that used by Pechacek et al. (1984). Students were told that the chemical saliva thiocyanate could be measured to determine the amount of their cigarette smoking. The researcher demonstrated placement of the cotton roll between the lower outside gum and the inside of the cheek, below the teeth. Once the cotton was in place, subjects began to fill out the questionnaire, following staff member instructions. At the end of the 8 minutes, students removed the cotton and continued working on the questionnaire. The completed assessment took 45 minutes. RESULTS
Incidence of self-reported marijuana and cigarette smoking Three groups were defined on the basis of self-reported marijuana and cigarette smoking in the previous week: (a) smokers of marijuana only, (b) smokers of cigarettes only, and (c) smokers of both marijuana and cigarettes. For each grade and the total sample, the proportion of subjects in each of these groups was examined together with their mean self-reported consumption of marijuana and/or cigarettes in the previous week. Smokers of marijuana alone made up 7.3% of the sample. There were 6.7% who smoked only cigarettes, and 6.1% who smoked both cigarettes and marijuana. Thus, 13.4% of the sample smoked marijuana and 12.8% smoked cigarettes. The proportions of subjects in each of these categories increased with increasing grade. The number of times that marijuana was smoked was not significantly higher in upper grades. The number of cigarettes consumed was significantly higher in higher grades, F(3,135) = 3.765, p < .012; according to an HSD test, rate of cigarette smoking in 10th grade was higher than in 7th, p < .Ol. Smokers who consumed both cigarettes and marijuana smoked significantly more marijuana than those who smoked only marijuana, F(1,142) = 13.58, p < .OOl. Moreover, they smoked more cigarettes than did those who smoked only cigarettes, F(1,135) = 8.03, p < .005. The correlation between cigarettes last week and marijuana last week was .408, p < .Ol. Correlations of self-reported smoking with CO The relationships of cigarette smoking, marijuana smoking, and grade level to CO were analyzed using multiple regression (Cohen & Cohen, 1975). For the purpose of these analyses, grade was contrast coded and entered into the equation as a set of three variables: (a) 7th versus 9th grade; (b) 7th and 9th versus IOth, 11th, and 12th grade; and (c) 10th versus 1lth and 12th grade. The score for cigarettes last week was entered
ANTHONY
140
BIGLAN
et al.
first, followed by marijuana in last week, and the set of three grade variables. Then the two-way interactions were entered: (a) interaction of cigarettes and marijuana in last week, (b) interaction of grade and cigarette smoking, and (c) interaction of grade and marijuana smoking. Finally, the three-way interaction of grade, cigarette smoking, and marijuana smoking was entered. Cigarette smoking and CO. Table 1 presents zero order and multiple correlations between self-reports of smoking and CO. The composite measure of cigarettes last week correlated .71 (p < .Ol) with CO. The size of the correlation between cigarettes last week and CO varied with grade. Multiple regression showed that the relationship between the number of cigarettes in the last week and CO was significantly stronger in the 1lth and 12th grade sample (r = .80) than it was in the 10th grade sample (r = .67), increment in R2 for this contrast = .058, F(1,1088) = 165.63, p < .Ol. The relationship was also stronger in the 9th grade (r = .75) than in the 7th grade (r = .5 l), increment in R2 for this contrast = .018, F(1,1088) = 52.35, p < .Ol. However, it should be noted that the correlation for 10th graders was actually lower than that for 9th graders. In a separate analysis this difference was found to be significant (p < .Ol). The data were examined to see if differences among grades in these correlations could be due to one of three artifacts. First, in order to assess whether differences in the proportion of smokers in each grade accounted for differences in the correlation, the analysis was repeated using smokers only. The correlation for the 9th graders (r = .66) Table
1.
Multiple
regression
Variable Cigarettes in Last Week Marijuana in Last Week Grade 7 vs. Grade 9 Grades 7 & 9 vs. lo-12 Grade 10 vs. 11 & 12
analysis
of cigarette,
AN Cigarette/Grade Effects
<
df
R
R’
,713 ,357 .I12 ,150 .074
,713 .716 .719 .723 .725
.508 .513 .517 .522 .526
.508 .005 .003 ,005 ,003
1450.17** 15.04** 9.51** 15.84** 9.59**
1,1088 1,1088 1,1088 1,1088 1,1088
.725
.526
.013
11.65**
3,1088
,528 ,372
,725 ,738
.526 ,544
.OOO .018
1.80 52.35**
1.1088 1,1088
.087
.738
,545
.ooo
.318
.776
,603
.058
165.63**
1,1088
,776
,603
,077
72.98**
3,1088
.780
.609
5.72*
3,1088
,259
,785
.615
19.04**
1,1088
.I15
.787
.619
9.51**
1,1088
,347
.787
.619
.787
.619
1,1088
Interaction Interaction lOth-12th Interaction & 12th
All Cigarette/Marijuana/Grade Interaction Effects **p
R’
F of Increment R’
Interaction
Cigarette/Marijuana times 7th vs. 9th Cigarette/Marijuana times 7th & 9th vs. Cigarette/Marijuana times 10th vs. 11th
.05;
Increment
on CO (n = 1103).
Interaction
All Marijuana/Grade Effects
*p <
and grade
r
All Grade Effects Cigarette/Marijuana Interaction Cigarette/7th vs. 9th Interaction Cigarette/7th & 9th vs. lo-12th Interaction Cigarette/lOth vs. 11th & 12th Interaction
marijuana,
.Ol
I,1088 9.52*
1,1088
Measurement of marijuana and cigarette smoking
141
was still significantly larger than that for 7th graders (r = 64) and the correlation for 11th and 12th graders (r = .73) was significantly greater than the correlation for 10th graders (r = .62). However, these results were not completely consistent with the notion that the CO-smoking rate correlation increases with age since the correlation for 10th graders was significantly lower than that for the 7th or 9th grades. Second, in order to see if differences among grades in the proportion of high rate smokers influenced the correlations, the analyses were repeated using only high rate smokers. The correlation for 11th and 12th grades was still significantly greater than that for 10th grade. Finally, subjects who exaggerated how much smoking they were doing could have contributed to lower correlations in some grades. We eliminated subjects who reported smoking 20 or more cigarettes a week but had CO levels below 5 ppm and recomputed the correlations for smokers only. When this was done the 10th grade correlation was still significantly smaller than the one for 11th and 12th grades. In a separate analysis, not shown in Table 1, the score for marijuana smoking was entered first, followed by that of cigarette smoking. In this analysis, marijuana last week accounted for 13% of the variance in CO, and cigarette smoking produced a large and significant increment in the variance of CO that was accounted for, increment in R* = .386, F(1,1088) = 1102.73, p < .Ol. When all marijuana smokers were deleted from the sample, the correlation between cigarettes last week and CO became somewhat smaller, r = .61, p < .Ol. Marijuana last week and CO. CO and self-reported marijuana smoking were significantly correlated, r = .36, p < .Ol. However, when variance due to cigarette smoking was removed, the amount of variance that marijuana smoking accounted for was negligible, increment in RZ = 305, F(1,1088) = 15.04, p < .Ol. When all cigarette smokers were deleted from the sample, the CO-marijuana last week correlation was .24, p < .Ol. The interaction between marijuana smoking and cigarette smoking did not account for significant variance in CO. The strength of the relationship between marijuana and CO varied with grade. However, the increment in R2 due to this interaction was quite small (increment in R2 = .006) Grade. A small but significant proportion of the variance in CO is uniquely due to grade. The increment to R2 when grade is added to cigarette and marijuana smoking is .013, F(3,1088) = 11.65, p < .Ol. Each of the three individual contrasts were significant: CO levels are higher in 9th than in 7th, are higher in 1lth and 12th than in lOth, and are higher in 10th through 12th than in 7th to 9th grades. Thus, even when the variance due to cigarette and marijuana smoking is removed, CO levels are slightly higher in higher grades. Interaction of grade with cigarette and marijuana smoking. The size of this interaction was so small that it does not appear worthwhile to interpret it, increment in R2 = .009, F(3,1097) = 7.40, p < .Ol. Correlations of self-reported smoking wiith SCN Because SCN data were available from only some self-reported nonsmokers, the cases were weighted so that the total N for the correlations remained the same, but the weight given to nonsmokers was proportionate to their numbers in the total sample.
142
ANTHONY
BIGLAN
et al.
Cigarette smoking. Self-reported smoking of cigarettes correlated significantly with SCN, r = .30, p < .Ol. When marijuana smokers were deleted from the sample, the correlation was .19, p < .Ol . The strength of this relationship did not differ according to grade. Marijuana smoking. Self-reported marijuana smoking was not highly related with SCN, r = .09, p < .05. With cigarette smokers deleted from the sample, the correlation was - .Ol, n.s. When variance due to cigarette smoking was removed, marijuana smoking did not correlate with SCN. Neither the interaction of marijuana smoking with cigarette smoking nor the interaction of marijuana smoking with grade was significant. Grade. Grade accounted for a moderate amount of the variance in SCN after variance due to cigarette and marijuana smoking had been removed, increment in RZ = .08, F(1,441) = 15.31, p < .Ol. This was primarily due to lower levels of SCN in 7th and 9th grades than in 10th through 12th grades. Interaction of grade, cigarette, and marijuana smoking. This three-way
interaction
was not significant. DISCUSSION
Adolescents’ self-reported cigarette smoking is significantly correlated with CO even when there is a sizeable number of marijuana smokers in the sample. The main reason for this appears to be that marijuana and cigarette smoking are correlated. Excluding subjects who report smoking marijuana from the sample actually reduces the correlation between CO and self-reported cigarette smoking because many of the high rate marijuana smokers are also high rate cigarette smokers. An additional factor may be a difference in the temporal pattern of marijuana and cigarette smoking. If marijuana were smoked primarily after school and on weekends and cigarette smoking also occurred prior to and during the school day, marijuana smoking would have less effect on CO levels measured during the school day than would cigarette smoking. The validity of the CO measure in assessing cigarette smoking could be jeopardized in a population in which marijuana smoking and cigarette smoking were not highly correlated. For example, Dull and Williams (198 1) found that the relationship between self-reported cigarette and marijuana smoking was greater in a sample of adolescents than it was in a sample of adults. The present study is consistent with Pechacek et al. (1984) in finding that CO predicts self-reported cigarette smoking better than does SCN. However, in the present study, the correlation between SCN and self-reported cigarette smoking (r = .30) was a good deal smaller than that (r = .48) found by Pechacek et al. (1984). The difference could not be due to variations in laboratory procedure, since the Minnesota Laboratory processed SCN for both studies. The length and/or method of storage of our SCN may have led to deterioration of our samples (Prue, Martin, Hume, & Davis, 1981). If this speculation is correct, the present study may underestimate the correlation between SCN and self-reported marijuana smoking. There was evidence of higher SCN and CO levels in upper grades even after the variance due to self-reported cigarette and marijuana smoking was removed. The size of these effects was small. However, they do suggest that some adolescents who smoke cigarettes and/or marijuana are denying that they do so. The higher correlations between self-reports of cigarette smoking and these physiological measures suggest that it
Measurement of marijuana and cigarette smoking
143
was denial of cigarette smoking rather than marijuana smoking that was the more important factor. The present results do not support the validity of expired air CO and saliva thiocyanate as measures of marijuana smoking in in-school assessments. The correlation of SCN with subjects’ reports of marijuana smoking was negligible. CO did correlate with reports of marijuana smoking, but the correlation was much smaller than that found between CO and cigarette smoking. It appears to be largely due to the co-occurrence of cigarette and marijuana smoking in this sample. When the variance due to self-reported cigarette smoking is removed from the CO-marijuana relationship, the relationship is negligible. Marijuana smoke does contain CO (Rickert, Robinson, & Rogers, 1982). Direct measurement of marijuana smoking topography and the CO boost that it produces are needed to clarify the relationship between marijuana smoking and CO. Grade and the dose per cigarette The present study partially replicates the finding of Pechacek et al. (1984) that the correlation between CO and self-reported cigarette smoking is higher with increasing grade. It does not replicate such an increasing correlation for SCN, but this may be due to inaccuracies in SCN assessment. The correlation between CO and self-reported smoking was largest for 1lth and 12th graders and smallest for 7th graders.,However, the correlation for 9th graders was larger than the one for 10th graders. This pattern remained even when the correlations were recomputed for (a) smokers only, (b) higher rate smokers only, and (c) all self-reported smokers except those who appeared to have exaggerated their smoking. This pattern may reflect differences among grade cohorts in the dose of CO they receive per cigarette. It is plausible that younger students generally puff less frequently on the cigarettes they smoke. However, the fact that the 10th grade correlation is lower than the one for 9th grade is not consistent with this notion. Direct assessments of adolescent smokers’ topography are needed to confirm whether such differences in dosage occur. An alternative explanation is that the differences reflect grade cohort differences in the accuracy of reporting smoking rate. Here, too, the correlation for 10th grade is an anomoly. The relationship between marijuana and cigarette smoking Subjects who report smoking both cigarettes and marijuana smoke more cigarettes than other cigarette smokers and more marijuana than other marijuana smokers. Among groups where such a relationship exists it would be more difficult to assess the health consequences of marijuana smoking. The possible contribution of cigarette smoking to correlations between illness and marijuana smoking would need to be examined in such studies. REFERENCES Bauman, K.E., & Dent, C.W. (1982). Influence of an objective measure of self-reports of behavior. Journal of Applied Psychology, 67, 623-628. Cohen, J., & Cohen, P. (1975). Applied multiple regression/correlation analysis for the behavioralsciences. New York: Halstead Press. Dull, R.T., & Williams, F.P., III. (1981). Marijuana, alcohol and tobacco: Reassessment of a presumed relationship. Journal of Drug Education, 11, 129-139. Evans, R.I., Hansen, W.B., & Mittlemark, M.B. (1977). Increasing the validity of self-reports of smoking behavior in children. Journal of Applied Psychology, 62, 521-523. Luepker, R.V., Pechacek, T.F., Murray, D.&l., Johnson, C.A., Hurd, F., & Jacobs, D.R. (1981). Saliva thiocyanate: A chemical indicator of cigarette smoking in adolescents. American Journal of Public Health, 71, 1320-1324. Pechacek, T.F., Murray, D.M., Luepker, R.V., Mittelmark, M.B., Johnson, C.A., & Shultz, B.A.
144
ANTHONY BIGLAN et al.
(1984). Measurement of adolescent smoking behavior: Rationale and methods. Journal Medicine,
of Behavioral
7, 123-140.
Penning, M., & Barnes, G.E. (1982). Adolescent marijuana use: A review. International Journal of the Addictions,
II, 749-791.
Prue, D.M., Martin, J.E., Hume, A.S., & Davis, N.S. (1981). The reliability of thiocyanate measurement of smoking exposure. Addictive Behaviors, 6, 99-105. Rickert, W.S., Robinson, J.C., & Rogers, B. (1982). A comparison of tar, carbon monoxide and pH levels in smoke from marijuana and tobacco cigarettes. Cunudiun Journal of Public Health, 73, 386-391.
0306-4603185 $3.00 + .OO Copyright e 1985 Pergamon Press Ltd
EXPIRED AIR CARBON MONOXIDE AND SALIVA THIOCYANATE: RELATIONSHIPS TO SELF-REPORTS OF MARIJUANA AND CIGARETTE SMOKING ANTHONY
BIGLAN, CHERI GALLISON, ROBERT THOMPSON
DENNIS ARY, and
Oregon Research Institute Abstract-This study examined the relationship between self-reports of marijuana and cigarette smoking, and the physiological measures of expired air carbon monoxide (CO) and saliva thiocyanate (SCN) in a sample of 1,130 seventh, ninth, tenth, eleventh, and twelfth graders. Subjects who reported marijuana smoking were likely to also report cigarette smoking. The correlation between self-reported marijuana smoking and SCN was negligible. The correlation between CO and self-reported marijuana smoking was statistically significant, but when the variance due to cigarette smoking was removed, this relationship also became negligible. The existence of a sizeable number of marijuana smokers in this sample of adolescents did not alter the correlation between CO and self-reports of cigarette smoking. However, in adult samples, where marijuana and cigarette smoking are less highly correlated, marijuana smoking could affect the relationship between CO and self-reported cigarette smoking. Carbon monoxide predicted self-reported cigarette smoking better than did saliva thiocyanate. There was an interaction between grade and the CO/cigarette smoking correlation. The correlations were generally higher in upper grades.
This article examines the relationships between two physiological measures and adolescents’ self-reports of marijuana and cigarette smoking in a sample of 7th, 9th, lOth, llth, and 12th graders. Expired air carbon monoxide (CO) and saliva thiocyanate (SCN) have been shown to correlate with self-reports of adolescent cigarette smoking (Pechacek, Murray, Luepker, Mittelmark, Johnson, & Shultz, 1984). However, the possible impact of marijuana on these measures has not been examined among subjects who routinely smoke marijuana. The CO and SCN levels associated with self-reported marijuana smoking are of interest for two reasons. First, if there are strong correlations between reports of marijuana smoking and physiological measures, the measures could be useful in corroborating reports of marijuana smoking, especially among persons who do not smoke cigarettes. Second, because marijuana smoking appears to occur among a significant proportion of young people (Penning & Barnes, 1982), it may affect the validity of these physiological measures as assessments of adolescent cigarette smoking. Thus, in the present study the relationship between self-reported marijuana and cigarette smoking is examined and the effect of marijuana smoking on the correlations between cigarette smoking and CO and SCN are assessed. Pechacek et al. (1984) have presented evidence suggesting that the relationships between SCN and CO and the number of cigarettes smoked is greater with increasing age. This implies that younger smokers may receive a smaller dose per cigarette and that Portions of this paper were presented at the American Psychological Association, Anaheim, 1983. This study was supported by Grant #HD 13409 and Grant #HD 15825 from the National Institute of Child Health and Human Development. The authors wish to thank Agatha McLean for typing assistance and Dayna Mitchell, LaVonne Boss, and Kathy Fisher for assistance in data collection. Requests for reprints should be sent to Anthony Biglan, Oregon Research Institute, 195 West 12th Ave., Eugene, OR 97401. 137
138
ANTHONY
BIGLAN
et al.
adolescents may take longer to develop adult-type inhalation patterns than was previously assumed. In the present study, the replicability of the Pechacek et al. (1984) finding is assessed and the following rival explanations are examined empirically. First, the results may have been due to the smaller proportion of smokers in lower grades and the fact that smoking rates among younger smokers were more restricted. Both of these factors would tend to reduce the correlation between self-reported cigarette smoking and CO or SCN. Second, Pechacek et al. (1984) identified several subjects in the lower age groups who may have been exaggerating how much they smoked. This would produce lower correlations.
METHOD
Subjects Subjects were 1,130 students who were participating in an experimental evaluation of a smoking prevention project. They were distributed across grades as follows: 455 seventh graders, 346 ninth graders, 150 tenth graders, and 179 eleventh and twelfth graders. The complete sample included 573 females and 557 males. The majority of the subjects analyzed (87%) were being assessed at 6-month follow-up with (13%) assessed prior to any intervention. The sample included three high schools and four middle schools. Measures Expired air carbon monoxide. The expired air sample was assessed for CO content using an Energetic Science Ecolyzer 2000. The assessment occurred within 12 hours of sample collection. Saliva thiocyanate (SCN). Saliva samples were collected in a sterile Falcon 17 x 100, 16 mm test tube using 1%” sterile dental cotton rolls. Saliva samples were stored within eight hours of collection in a professional freezer locker at - .18”F. For analysis, they were shipped frozen to the University of Minnesota Health Behaviors Laboratory where saliva thiocyanate levels were determined by spectophotometer (Luepker, Pechacek, Murry, Johnson, Hurd, & Jacobs, 1981). Because of the expense of analyzing saliva thiocyanate samples, it was necessary to sample subjects on this variable. All self-reported smokers of cigarettes and/or marijuana were selected for analysis from the total sample along with 25% of the self-reported nonsmoking subjects. The samples were stored for 2 to 8 months before shipping. Questionnaire. Current self-reported smoking was assessed by asking subjects to write down the number of cigarettes they smoked in the last week and the number of cigarettes they smoked in the previous 24 hours. A weighted average of these items(Cigarettes last week + 7 x cigarettes yesterday)/2 - was composed. Marijuana smoking was assessed by asking subjects to indicate the number of times they smoked marijuana in the last week. Subjects could choose a number 0 through 10 or the item “more than ten times.” For the purposes of analysis, the latter response was coded as eleven. Procedures Two weeks prior to data collection, a researcher described the project in each classroom that was to be assessed. Students were told that the assessment consisted of completing a questionnaire regarding their smoking experiences, and supplying breath and saliva samples. Simultaneously, a letter was mailed to the parents, explaining the as-
Measurement of marijuana and cigarette smoking
139
sessment for which their child was volunteering. Parents who did not want their child to participate returned a postcard indicating their declination. Two or three staff members of the smoking prevention project collected the data in the classroom setting. Subjects began by filling out an identification sheet. Following this, they provided an expired air carbon monoxide (CO) sample. Prior to collection of the breath sample, subjects were informed that analysis of expired air for carbon monoxide (CO) content would allow identification of cigarette smokers. This was done because there is evidence that such instructions increase the accuracy of self-reports of cigarette smoking (Evans, Hansen, & Mittlemark, 1977; Bauman & Dent, 1982). The air sample collection procedure involved subjects holding their breath for 15 seconds without taking in additional air. After 15 seconds, subjects breathed into a l- or 2-liter polyethylene bag, then securely sealed it with a rubberband. Next, subjects provided a saliva sample using a procedure similar to that used by Pechacek et al. (1984). Students were told that the chemical saliva thiocyanate could be measured to determine the amount of their cigarette smoking. The researcher demonstrated placement of the cotton roll between the lower outside gum and the inside of the cheek, below the teeth. Once the cotton was in place, subjects began to fill out the questionnaire, following staff member instructions. At the end of the 8 minutes, students removed the cotton and continued working on the questionnaire. The completed assessment took 45 minutes. RESULTS
Incidence of self-reported marijuana and cigarette smoking Three groups were defined on the basis of self-reported marijuana and cigarette smoking in the previous week: (a) smokers of marijuana only, (b) smokers of cigarettes only, and (c) smokers of both marijuana and cigarettes. For each grade and the total sample, the proportion of subjects in each of these groups was examined together with their mean self-reported consumption of marijuana and/or cigarettes in the previous week. Smokers of marijuana alone made up 7.3% of the sample. There were 6.7% who smoked only cigarettes, and 6.1% who smoked both cigarettes and marijuana. Thus, 13.4% of the sample smoked marijuana and 12.8% smoked cigarettes. The proportions of subjects in each of these categories increased with increasing grade. The number of times that marijuana was smoked was not significantly higher in upper grades. The number of cigarettes consumed was significantly higher in higher grades, F(3,135) = 3.765, p < .012; according to an HSD test, rate of cigarette smoking in 10th grade was higher than in 7th, p < .Ol. Smokers who consumed both cigarettes and marijuana smoked significantly more marijuana than those who smoked only marijuana, F(1,142) = 13.58, p < .OOl. Moreover, they smoked more cigarettes than did those who smoked only cigarettes, F(1,135) = 8.03, p < .005. The correlation between cigarettes last week and marijuana last week was .408, p < .Ol. Correlations of self-reported smoking with CO The relationships of cigarette smoking, marijuana smoking, and grade level to CO were analyzed using multiple regression (Cohen & Cohen, 1975). For the purpose of these analyses, grade was contrast coded and entered into the equation as a set of three variables: (a) 7th versus 9th grade; (b) 7th and 9th versus IOth, 11th, and 12th grade; and (c) 10th versus 1lth and 12th grade. The score for cigarettes last week was entered
ANTHONY
140
BIGLAN
et al.
first, followed by marijuana in last week, and the set of three grade variables. Then the two-way interactions were entered: (a) interaction of cigarettes and marijuana in last week, (b) interaction of grade and cigarette smoking, and (c) interaction of grade and marijuana smoking. Finally, the three-way interaction of grade, cigarette smoking, and marijuana smoking was entered. Cigarette smoking and CO. Table 1 presents zero order and multiple correlations between self-reports of smoking and CO. The composite measure of cigarettes last week correlated .71 (p < .Ol) with CO. The size of the correlation between cigarettes last week and CO varied with grade. Multiple regression showed that the relationship between the number of cigarettes in the last week and CO was significantly stronger in the 1lth and 12th grade sample (r = .80) than it was in the 10th grade sample (r = .67), increment in R2 for this contrast = .058, F(1,1088) = 165.63, p < .Ol. The relationship was also stronger in the 9th grade (r = .75) than in the 7th grade (r = .5 l), increment in R2 for this contrast = .018, F(1,1088) = 52.35, p < .Ol. However, it should be noted that the correlation for 10th graders was actually lower than that for 9th graders. In a separate analysis this difference was found to be significant (p < .Ol). The data were examined to see if differences among grades in these correlations could be due to one of three artifacts. First, in order to assess whether differences in the proportion of smokers in each grade accounted for differences in the correlation, the analysis was repeated using smokers only. The correlation for the 9th graders (r = .66) Table
1.
Multiple
regression
Variable Cigarettes in Last Week Marijuana in Last Week Grade 7 vs. Grade 9 Grades 7 & 9 vs. lo-12 Grade 10 vs. 11 & 12
analysis
of cigarette,
AN Cigarette/Grade Effects
<
df
R
R’
,713 ,357 .I12 ,150 .074
,713 .716 .719 .723 .725
.508 .513 .517 .522 .526
.508 .005 .003 ,005 ,003
1450.17** 15.04** 9.51** 15.84** 9.59**
1,1088 1,1088 1,1088 1,1088 1,1088
.725
.526
.013
11.65**
3,1088
,528 ,372
,725 ,738
.526 ,544
.OOO .018
1.80 52.35**
1.1088 1,1088
.087
.738
,545
.ooo
.318
.776
,603
.058
165.63**
1,1088
,776
,603
,077
72.98**
3,1088
.780
.609
5.72*
3,1088
,259
,785
.615
19.04**
1,1088
.I15
.787
.619
9.51**
1,1088
,347
.787
.619
.787
.619
1,1088
Interaction Interaction lOth-12th Interaction & 12th
All Cigarette/Marijuana/Grade Interaction Effects **p
R’
F of Increment R’
Interaction
Cigarette/Marijuana times 7th vs. 9th Cigarette/Marijuana times 7th & 9th vs. Cigarette/Marijuana times 10th vs. 11th
.05;
Increment
on CO (n = 1103).
Interaction
All Marijuana/Grade Effects
*p <
and grade
r
All Grade Effects Cigarette/Marijuana Interaction Cigarette/7th vs. 9th Interaction Cigarette/7th & 9th vs. lo-12th Interaction Cigarette/lOth vs. 11th & 12th Interaction
marijuana,
.Ol
I,1088 9.52*
1,1088
Measurement of marijuana and cigarette smoking
141
was still significantly larger than that for 7th graders (r = 64) and the correlation for 11th and 12th graders (r = .73) was significantly greater than the correlation for 10th graders (r = .62). However, these results were not completely consistent with the notion that the CO-smoking rate correlation increases with age since the correlation for 10th graders was significantly lower than that for the 7th or 9th grades. Second, in order to see if differences among grades in the proportion of high rate smokers influenced the correlations, the analyses were repeated using only high rate smokers. The correlation for 11th and 12th grades was still significantly greater than that for 10th grade. Finally, subjects who exaggerated how much smoking they were doing could have contributed to lower correlations in some grades. We eliminated subjects who reported smoking 20 or more cigarettes a week but had CO levels below 5 ppm and recomputed the correlations for smokers only. When this was done the 10th grade correlation was still significantly smaller than the one for 11th and 12th grades. In a separate analysis, not shown in Table 1, the score for marijuana smoking was entered first, followed by that of cigarette smoking. In this analysis, marijuana last week accounted for 13% of the variance in CO, and cigarette smoking produced a large and significant increment in the variance of CO that was accounted for, increment in R* = .386, F(1,1088) = 1102.73, p < .Ol. When all marijuana smokers were deleted from the sample, the correlation between cigarettes last week and CO became somewhat smaller, r = .61, p < .Ol. Marijuana last week and CO. CO and self-reported marijuana smoking were significantly correlated, r = .36, p < .Ol. However, when variance due to cigarette smoking was removed, the amount of variance that marijuana smoking accounted for was negligible, increment in RZ = 305, F(1,1088) = 15.04, p < .Ol. When all cigarette smokers were deleted from the sample, the CO-marijuana last week correlation was .24, p < .Ol. The interaction between marijuana smoking and cigarette smoking did not account for significant variance in CO. The strength of the relationship between marijuana and CO varied with grade. However, the increment in R2 due to this interaction was quite small (increment in R2 = .006) Grade. A small but significant proportion of the variance in CO is uniquely due to grade. The increment to R2 when grade is added to cigarette and marijuana smoking is .013, F(3,1088) = 11.65, p < .Ol. Each of the three individual contrasts were significant: CO levels are higher in 9th than in 7th, are higher in 1lth and 12th than in lOth, and are higher in 10th through 12th than in 7th to 9th grades. Thus, even when the variance due to cigarette and marijuana smoking is removed, CO levels are slightly higher in higher grades. Interaction of grade with cigarette and marijuana smoking. The size of this interaction was so small that it does not appear worthwhile to interpret it, increment in R2 = .009, F(3,1097) = 7.40, p < .Ol. Correlations of self-reported smoking wiith SCN Because SCN data were available from only some self-reported nonsmokers, the cases were weighted so that the total N for the correlations remained the same, but the weight given to nonsmokers was proportionate to their numbers in the total sample.
142
ANTHONY
BIGLAN
et al.
Cigarette smoking. Self-reported smoking of cigarettes correlated significantly with SCN, r = .30, p < .Ol. When marijuana smokers were deleted from the sample, the correlation was .19, p < .Ol . The strength of this relationship did not differ according to grade. Marijuana smoking. Self-reported marijuana smoking was not highly related with SCN, r = .09, p < .05. With cigarette smokers deleted from the sample, the correlation was - .Ol, n.s. When variance due to cigarette smoking was removed, marijuana smoking did not correlate with SCN. Neither the interaction of marijuana smoking with cigarette smoking nor the interaction of marijuana smoking with grade was significant. Grade. Grade accounted for a moderate amount of the variance in SCN after variance due to cigarette and marijuana smoking had been removed, increment in RZ = .08, F(1,441) = 15.31, p < .Ol. This was primarily due to lower levels of SCN in 7th and 9th grades than in 10th through 12th grades. Interaction of grade, cigarette, and marijuana smoking. This three-way
interaction
was not significant. DISCUSSION
Adolescents’ self-reported cigarette smoking is significantly correlated with CO even when there is a sizeable number of marijuana smokers in the sample. The main reason for this appears to be that marijuana and cigarette smoking are correlated. Excluding subjects who report smoking marijuana from the sample actually reduces the correlation between CO and self-reported cigarette smoking because many of the high rate marijuana smokers are also high rate cigarette smokers. An additional factor may be a difference in the temporal pattern of marijuana and cigarette smoking. If marijuana were smoked primarily after school and on weekends and cigarette smoking also occurred prior to and during the school day, marijuana smoking would have less effect on CO levels measured during the school day than would cigarette smoking. The validity of the CO measure in assessing cigarette smoking could be jeopardized in a population in which marijuana smoking and cigarette smoking were not highly correlated. For example, Dull and Williams (198 1) found that the relationship between self-reported cigarette and marijuana smoking was greater in a sample of adolescents than it was in a sample of adults. The present study is consistent with Pechacek et al. (1984) in finding that CO predicts self-reported cigarette smoking better than does SCN. However, in the present study, the correlation between SCN and self-reported cigarette smoking (r = .30) was a good deal smaller than that (r = .48) found by Pechacek et al. (1984). The difference could not be due to variations in laboratory procedure, since the Minnesota Laboratory processed SCN for both studies. The length and/or method of storage of our SCN may have led to deterioration of our samples (Prue, Martin, Hume, & Davis, 1981). If this speculation is correct, the present study may underestimate the correlation between SCN and self-reported marijuana smoking. There was evidence of higher SCN and CO levels in upper grades even after the variance due to self-reported cigarette and marijuana smoking was removed. The size of these effects was small. However, they do suggest that some adolescents who smoke cigarettes and/or marijuana are denying that they do so. The higher correlations between self-reports of cigarette smoking and these physiological measures suggest that it
Measurement of marijuana and cigarette smoking
143
was denial of cigarette smoking rather than marijuana smoking that was the more important factor. The present results do not support the validity of expired air CO and saliva thiocyanate as measures of marijuana smoking in in-school assessments. The correlation of SCN with subjects’ reports of marijuana smoking was negligible. CO did correlate with reports of marijuana smoking, but the correlation was much smaller than that found between CO and cigarette smoking. It appears to be largely due to the co-occurrence of cigarette and marijuana smoking in this sample. When the variance due to self-reported cigarette smoking is removed from the CO-marijuana relationship, the relationship is negligible. Marijuana smoke does contain CO (Rickert, Robinson, & Rogers, 1982). Direct measurement of marijuana smoking topography and the CO boost that it produces are needed to clarify the relationship between marijuana smoking and CO. Grade and the dose per cigarette The present study partially replicates the finding of Pechacek et al. (1984) that the correlation between CO and self-reported cigarette smoking is higher with increasing grade. It does not replicate such an increasing correlation for SCN, but this may be due to inaccuracies in SCN assessment. The correlation between CO and self-reported smoking was largest for 1lth and 12th graders and smallest for 7th graders.,However, the correlation for 9th graders was larger than the one for 10th graders. This pattern remained even when the correlations were recomputed for (a) smokers only, (b) higher rate smokers only, and (c) all self-reported smokers except those who appeared to have exaggerated their smoking. This pattern may reflect differences among grade cohorts in the dose of CO they receive per cigarette. It is plausible that younger students generally puff less frequently on the cigarettes they smoke. However, the fact that the 10th grade correlation is lower than the one for 9th grade is not consistent with this notion. Direct assessments of adolescent smokers’ topography are needed to confirm whether such differences in dosage occur. An alternative explanation is that the differences reflect grade cohort differences in the accuracy of reporting smoking rate. Here, too, the correlation for 10th grade is an anomoly. The relationship between marijuana and cigarette smoking Subjects who report smoking both cigarettes and marijuana smoke more cigarettes than other cigarette smokers and more marijuana than other marijuana smokers. Among groups where such a relationship exists it would be more difficult to assess the health consequences of marijuana smoking. The possible contribution of cigarette smoking to correlations between illness and marijuana smoking would need to be examined in such studies. REFERENCES Bauman, K.E., & Dent, C.W. (1982). Influence of an objective measure of self-reports of behavior. Journal of Applied Psychology, 67, 623-628. Cohen, J., & Cohen, P. (1975). Applied multiple regression/correlation analysis for the behavioralsciences. New York: Halstead Press. Dull, R.T., & Williams, F.P., III. (1981). Marijuana, alcohol and tobacco: Reassessment of a presumed relationship. Journal of Drug Education, 11, 129-139. Evans, R.I., Hansen, W.B., & Mittlemark, M.B. (1977). Increasing the validity of self-reports of smoking behavior in children. Journal of Applied Psychology, 62, 521-523. Luepker, R.V., Pechacek, T.F., Murray, D.&l., Johnson, C.A., Hurd, F., & Jacobs, D.R. (1981). Saliva thiocyanate: A chemical indicator of cigarette smoking in adolescents. American Journal of Public Health, 71, 1320-1324. Pechacek, T.F., Murray, D.M., Luepker, R.V., Mittelmark, M.B., Johnson, C.A., & Shultz, B.A.
144
ANTHONY BIGLAN et al.
(1984). Measurement of adolescent smoking behavior: Rationale and methods. Journal Medicine,
of Behavioral
7, 123-140.
Penning, M., & Barnes, G.E. (1982). Adolescent marijuana use: A review. International Journal of the Addictions,
II, 749-791.
Prue, D.M., Martin, J.E., Hume, A.S., & Davis, N.S. (1981). The reliability of thiocyanate measurement of smoking exposure. Addictive Behaviors, 6, 99-105. Rickert, W.S., Robinson, J.C., & Rogers, B. (1982). A comparison of tar, carbon monoxide and pH levels in smoke from marijuana and tobacco cigarettes. Cunudiun Journal of Public Health, 73, 386-391.