Comparison of Tests Used to Distinguish Smokers from Non-Smokers

American Journal of Public Health Reprint.

Martin J. Jarvis, MPhil, Hugh Tunstall-Pedoe, MD, Colin Feyeraband, PhD, Cyril Vesey, MSc, and Yussuf Saloojee, MIBiol

ABSTRACT: Questionnaire and biochemical measures of smoking in 211 hospital outpatients. Eleven different tests of smoke intake were compared for their ability to categorize smokers and nonsmokers correctly. The concentration of cotinine, whether measured in plasma, saliva, or urine, was the best indicator of smoking, with sensitivity of 96-97 percent and specificity of 99-100 percent. Thiocyanate provided the poorest discrimination. Carbon monoxide measured as blood carboxyhaemoglobin or in expired air gave sensitivity and specificity of about 90 percent. Sensitivities of the tests were little affected by the presence among the claimed nonsmokers of a group of 21 “deceivers” who concealed their smoking. It is concluded that cotinine is the measure of choice, but for most clinical applications carbon monoxide provides and acceptable degree of discrimination and is considerably cheaper and simpler to apply. (Am. J Public Health 1987; 77: 1435-1438)


Self-reports of smoking status may not always be reliable, particularly in situations where smokers feel under strong pressure to give up smoking but have not been able to achieve this.1-3 A number of biochemical markers have been used to validate claims of nonsmoking, including measures based on thiocyanate, 4-7 nicotine, 8 cotinine, 9-11 and carbon monoxide.4, 6, 11, 12 These measures differ widely in availability, cost, and ease of administration. Measures based on nicotine have the advantage of being specific to tobacco but require expensive laboratory instrumentation. Levels of thiocyanate and carbon monoxide are much easier to determine but may be raised through exposures unrelated to smoking, such as traffic emissions and diet. Few studies have attempted to compare the various biochemical tests. 11 We report here a study in which all the markers of smoking currently in widespread use are all compared for their ability to categorize smokers and nonsmokers correctly.



The subjects for the study were a group of 215 outpatients at St. Mary’s Hospital, London. On arrival for their clinic appointment, they were asked to fill in a self-completion questionnaire giving details of smoking habits and to provide samples of blood, expired air, saliva, and urine. There was no prior warning of the survey, but consent for the biochemical tests was obtained before completion of the questionnaire. It was emphasized that the questionnaire responses were confidential and would not become part of hospital notes or be communicated to the medical staff. It was hoped that this would encourage accurate self-report. The present report is confined to 211 subjects who provided adequate questionnaire and biochemical data. There were 159 men (average age 56.0) and 52 women (average age 55.3); 119 attended afternoon cardiology clinics, and 92 a morning peripheral vascular clinic. A high proportion were suffering from smoking-related diseases. A total of 188 (89 percent) reported having been cigarette smokers at some time and 90 (43 percent) said that they were current smokers of cigarettes, pipes, or cigars. Reported mean cigarette consumption in the cigarette smokers was 13.2 cigarettes per day, and 97 percent reported having smoked on the test day, with a mean time since last cigarette of 1.5 hours.

The concentration of nicotine and cotinine in plasma, saliva, and urine was determined by gas chromatography.13, 14 Carboxyhaemoglobin concentrations were measured with an IL282 CO-Oximeter and carbon monoxide in expired after breath holding with a portable CO analyzer incorporating an ethanol filter.15 Thiocyanate was measured by an automated modification of the Aldridge technique.16 Urinary concentrations of nicotine, cotinine, and thiocyanate were not adjusted for urine flow.

Self Reported Smoking Categories
Cigarette smokers were defined as those who answered “yes” to the question “Do you smoke cigarettes now?” Pipe and cigar smokers answered “no” to this questions but “yes” to “Do you smoke a pipe?” or “Do you smoke cigars?” Nonsmokers were those who answered “no” to all these questions.

The sensitivity (percent of smokers detected) and specificity (percent of nonsmokers correctly classified) of each test was first examined by selecting the value which, in relation to self reported smoking, misclassified the fewest subjects. Subsequent analysis of sensitivity and specificity made allowance for claimed nonsmoking in active smokers (“deceivers”). Finally, a comparison between the tests was made by examining their sensitivity when specificity was standardized at 95 percent.17


item_1 The mean value for each marker are shown in Table 1 according to self-reported smoking status. As would be expected, concentrations both in the cigarette smokers and in those who smoked cigars or pipes but not cigarettes were greatly raised by comparison with the claimed nonsmokers. Table 2 gives the optimal cut-off value, sensitivity and specificity for each marker in relation to self-reported smoking status. With the exception of urinary thiocyanate, all the tests performed reasonably well in identifying smokers. Sensitivities were higher for measures based on nicotine than for CO or thiocyanate, and were slightly higher for identifying cigarette smoking rather than smoking of any tobacco product. Measures of cotinine correctly classified all but one or two of the self-reported smokers.


Specificity was lass satisfactory. About 15-20 percent of claimed nonsmokers were classified as smokers by each test. For the measures of nicotine and cotinine this would be unexpected, since nicotine is specific to tobacco and passive exposure to other people’s smoke has been found to have only a small effect in raising concentrations.18 It therefore seemed probable that there was some denial of active smoking among the claimed nonsmokers. Inspections of the data for individual subjects lent support to this. The figure shows the distribution of values of plasma cotinine according to self-reported smoking. The distribution in the claimed nonsmokers was clearly bimodal, with 21 subjects having concentrations similar to those seen in the smokers and two orders of magnitude higher than in the rest of the nonsmokers. This pattern was repeated across the other nicotine-based measures, and also recurred with the non-nicotine tests. For example, 18 out of 21 “apparent deceivers” had expired air CO concentrations which exceeded the nonsmoker-smoker cut-off, and in 16 out of 21 the concentration of plasma thiocyanate was similarly raised.


Table 3 shows mean values for each marker for true nonsmokers and for the 21 “deceivers”. In each case the concentration in the deceivers was raised substantially by comparison with “true” nonsmokers. For 8 out of 11 markers there was no real difference between “deceivers” and self-reported smokers. In subsequent analyses, the deceivers were therefore combined with the other smokers to form a true smoking group (n=111).

Table 4 gives the cut-off, sensitivity and specificity for each marker in discriminating true smoking. Cut-offs were unchanged from self-report for all measures except carbon monoxide. Specificities were markedly improved, and approached 100 percent for the nicotine based measures.

Comparison Between Markers item_4
In order to standardize comparison between tests, a cut-off for each was selected such that 95 percent of nonsmokers were correctly classified. Table 5 gives the corresponding sensitivities. Measures of thiocyanate in saliva and urine performed worst, with sensitivities of only 25 percent and 51 percent, respectively, for detecting cigarette smoking. Plasma thiocyanate fared better (74 percent), but still was substantially less sensitive than CO in expired air (88 percent) or blood carboxyhaemoglobin (89 percent). Measures based on nicotine showed the greatest sensitivity, with cotinine (99 percent) consistently better than nicotine itself (93-95 percent). All the tests were slightly less successful in detecting smoking of any tobacco product than in detecting cigarette smoking. Their failure to identify all smokers correctly reflected the presence in the sample of seven atypical light smokers. Four of these were cigarette smokers and 3 smoked cigars. They reported little or no inhalation, and most had not smoked on the day of the survey. Their marker concentrations were consistently at or near nonsmoker levels.


item_6 The results of this study indicate that the concentration in the body of certain biochemical markers of smoke intake can give a categorization of smoking status which is substantially more accurate than self-report. Cotinine, because of its longer half-life, performed better than nicotine. Since both are specific to tobacco, it is not surprising that they were more successful in identifying smokers than measures of carbon monoxide or thiocyanate. However, no measure can be 100 percent sensitive to active smoking or completely successful in identifying nonsmokers. This is because, on the one hand, exposure to other people’s smoke leads to the absorption of smoke products in nonsmokers and, on the other hand, because some smokers may smoke so infrequently or inhale so little that their intakes cannot reliably distinguishes from passive smoking.

Our results were robust and not influenced in any significant way by the presence in the sample of a substantial minority of smokers who denied their habit. When cut-offs were optimized to discriminate smoking according to self-report, the “deceivers” were correctly categorized as smokers, and their reallocation to the smoking group caused very little shift in cut-off values. The optimal cut-offs for the nicotine, cotinine, and thiocyanate measures all remained unchanged, and there were only minimal shifts in the cutting points for expired air carbon monoxide and COHb.


When biochemical markers are used to discriminate smoking status the optimal cut-point for any particular application will depend on the prevalence of smoking in the study population and on whether differential weights are attached to misclassifications of smokers and nonsmokers. Our cut-points were optimized for one particular sample, and the precise values may not therefore be directly generalizable to all clinical and research situations. However, the cut-points we found are similar to those reported by other workers for each of the markers studies, and may therefore provide an approximate indication of cut-off points likely to be useful in a variety of situations.

Our findings have implications for clinical practice in hospitals and other settings where patients may present with smoking-related disease. Self-reported smoking rates are likely to give a substantial underestimate of the true prevalence of smoking. Among our sample, 19 percent of smokers claimed to be nonsmokers. This misreporting could lead to an underestimate of the effects of smoking on the course of disease and could prejudicially affect decisions on patient management. At the same time, the success of objective tests in identifying smokers means that the clinician can escape from dependence on unreliable self-report.

Which then, of the tests examined, can be recommended for routine clinical use? Cotinine provides the best discrimination and must be the marker of choice for situations where accuracy is paramount. Non-invasive specimens of saliva or urine give essentially the same information as blood samples. But high costs and the elaborate laboratory facilities necessary argue against cotinine for routine use. The same is true for nicotine, where difficulties of avoiding sample contamination can raise added problems. Carbon monoxide levels provide a more sensitive indication of smoking than measure of thiocyanate and are cheaper and easier to apply. They are also likely to be valuable for excluding smoking in people who admit to using smokeless tobacco products and who therefore would have elevated level of nicotine and cotinine. The concentration of carbon monoxide in expired air can be measured no-invasively with a simple and inexpensive portable monitor by someone with minimal training, with essentially the same results as achieved by a more expensive and difficult assay requiring blood sample.19, 20 Although in principle smokers could avoid detection by a day’s abstinence before coming for a test, in practice this seems unlikely to occur, particularly when there is repeated and regular contact. In the present sample, 97 percent of self-reported cigarette smokers said they had smoked on the day of testing.

In conclusion, whether or not a person is a current smoker can be established accurately by objective tests of smoke intake. The few smokers who cannot be reliably identified smoke so infrequently or inhale so little that their habit is of minimal clinical significance. Although measures of cotinine perform best in discriminating smokers and nonsmokers and are the tests of choice for research protocols where accurate categorization is essential, for routine clinical applications expired air carbon monoxide is a simple, cheap, and acceptably accurate measure.


We thank the physicians responsible for cardiovascular clinics at St.Mary’s Hospital for their cooperation. Financial support was provided by the Medical Research Council and by the Joint Research of St. Bartholomew’s Hospital.


1. OhlinP, Lundh B, Westling H: Carbon monoxide blood levels and reported cessation of smoking. Psychopharmacology 1976; 46: 263-265.
2. Sillett RW, Wilson MB, Malcolm RE, Ball KP: Deception among smokers. Br Med J 1978; 2: 1185-1186.
3. Multiple Risk Factor Intervention Trial (MRFIT) Research Group: Multiple Risk Factor Intervention Trial: Risk factor changes and mortality results. JAMA 1982; 248: 1465-1477.
4. Vogt TM, Selvin S, Widdowson G, Hulley SB: Expired air carbon monoxide and serum thiocyanate as objective measure of cigarette exposure. AM J Public Health 1977; 67: 545-549.
5. Vesey CJ, Saloojee Y, Cole PV, Russell MAH. Blood carboxyhaemoglobin, plasma thiocyanate, and cigarette consumption: Implications for epidemiological studies in smokers. BR Med J 1982; 284: 1516-1518.
6. Saloojee Y, Vesey CJ, Cole PV, Russell MAH. Carboxyhaemoglobin and plasma thiocyanate: Complimentary indicators of smoking behavior. Thorax 1982; 37: 521-525.
7. Kornitzer M, Vanhemeldonck A, Bourdoux P, De Backer G. Belgian heart disease prevention project: Comparison of self-reported smoking behavior with serum thiocyanate concentrations. J Epidemiol Community Health 1983; 37: 132-136.
8. Wildox RG, Hughes J, Roland J. Verification of smoking history using urinary nicotine and cotinine measurements. BR Med J 1979; 2: 1026-1028.
9. Haley NJ, Axelrad MS, Tilton BS. Validation of self-reported smoking behavior: Biochemical analyses of cotinine and thiocyanane. Am J Public Health 1983; 2: 1204-1207.
10. Jamrozik K, Vesey M, Fowler G, Wald N, Parker G, Van Vunakis H. Controlled trial of three different antismoking interventions in general practice. Br Med J 1984; 288: 1499-1502.
11. Pojer R, Whitfield JB, Poulos V, Eckhart IF, Richmond R, Hensley WJ. Carboxyhaemoglobin, cotinine, and thiocyanate assay compared for distinguishing smoker from non-smokers. Clin Chem 1984; 30: 1377-1380.
12. Wald NJ, Idle M, Boreham J, Bailey A. Carbon monoxide in breath relation to smoking and carboxyhaemoglobin levels. Thorax 1981; 36:366-369.
13. Feyerabend C, Russell MAH. Assay of nicotine in biological materials sources of contaminationand their elimination. J Pharm Pharmacol 1980; 32: 178-181.
14. Feyerabend C, Russell MAH. A rapid gas-liquid chromatographic determination of cotinine in biological fluids. Analyst 1980; 105: 993-1001.
15. Jarvis MJ, Russell MAH, Saloojee Y. Expired air carbon monoxide: a simple breath test of tobacco smoke intake. BR Med J 1980; 281:484-485.
16. Vesey CJ, Kirk CJC. Two automated methods for measuring plasma thiocyanate compared. Clin Chem 1985; 31: 270-274.
17. Zweig MH, Robertson EA. Why we need better test evaluations. CLin Chem 1982; 28:1272-1276.
18. Jarvis MJ, Russell MAH, Feyerabend C, Eiser JR, Morgan M, Gammage P, Gray EM. Passive exposure to tobacco smoke: saliva cotinine concentrations in a representative population sample of nonsmoking school children. BR Med J 1985; 291: 927-929.
19. Jarvis MJ, Belcher M, Vesey C, Hutchison DCS. Low-cost carbon monoxide monitors in smoking assessment. Thorax 1986; 41: 886-887
20. Irving JM, Clark EC, Crombie IK, Smith WCS. Evaluation of a portable measure of expired carbon monoxide. Prev Me (In press).