Clinical examination
Although over-represented in primary health care settings, up to 90% of AOD misuse problems remain undetected by health professionals (Brown & Fleming, 1998; CASA, 2000; McPherson & Hersch, 2000; Roche et al., 2002). Unlike the somatic and often overt symptoms of alcohol abuse, identification of harmful and potentially harmful cannabis use presents a diagnostic challenge. Obstacles include denial or non-disclosure of cannabis use; no clear constellation of early signs specific to cannabis abuse; wide inter-individual variation in cannabis effects; likely distortion in subjective appraisal of related consequences; concurrent use of other drugs; fear of legal repercussions of
disclosure of illicit drug use, and difficulties in quantifying cannabis use. There are no guidelines on how to assess or differentiate dependent, harmful or at risk cannabis use (Carroll, 1995; Dennis et al., 2002b; Nathan, 1996; Roffman & Stephens, 1993; Swift et al., 1998a). Thus, cannabis problems may go undetected if less obvious physical symptoms (drug paraphernalia, trauma/accident history, disrupted sleep patterns, gastrointestinal disturbance, weight loss, poor nutritional intake, poor hygiene, respiratory infections/problems) and mental health symptoms (low self-esteem, restlessness and agitation, anger problems, anxiety and depression, suicidal ideation, unusual fears, memory deficits, difficulties in thinking and expressing thoughts, hallucinations or paranoia) are not routinely assessed (Carroll, 1995; Cottler & Compton, 1993). Those screening adolescents should be sensitive to “red flags” of serious cannabis-related problems (AACAP, 1997; McLellan & Dembo, 1993). “Red flags” include physical or sexual abuse; suicide attempts or other self-harm; parental drug abuse, dysfunctional family relationships; affiliation with delinquent peer group, social alienation, poor school/work attendance or performance; truancy, criminal offending, unprotected sex or other high-risk behaviours; and legal problems (AACAP, 1997; McLellan & Dembo, 1993; Monti et al., 2001; Spooner at al., 1996).
In sum, error and unreliability will seriously reduce the sensitivity of clinical examinations reliant on observation and medical history. Since clinically detectable symptoms tend to occur relatively late in the evolution of AOD problems, these procedures are unlikely to detect early-stage cannabis problems. A major practical limitation is that clinical screening requires the involvement of a primary health care practitioner.
Biochemical measures
Modern biochemical indicators of AOD use include urine, blood/plasma, scalp hair, saliva, fingernails, breath, tears, sweat, breast milk and meconium (Mura, Klintz, Papet, Ruesch, & Piriou, 1999; Wolff et al., 1999). Although less susceptible to many distortions, demand characteristics or biases associated with self-reports, these biological markers of cannabis use are not infallible (Carroll, 1995; Wolff et al., 1999). THC metabolites disappear rapidly from the bloodstream. After about 20 minutes, long
before the ‘high’ state has ended, no detectable metabolites remain. Laboratory error, misinterpretation, contamination, rapid decay, accidental or intentional donor dilution, adulteration or substitution of a freshly-voided sample, passive cannabis exposure, or concurrent use of other substances - all affect the accuracy of blood and urine tests (AACAP, 1997; Buchan, Dennis, Tims, & Diamond, 2002; Wolff et al., 1999).
Laboratory detection of targeted substances is complex, dependent on dose and pharmacokinetics of the drug ingested (Dawe et al., 2002; Wolff et al, 1999). Since drugs metabolise at different rates, urine testing cannot be used to determine dosage, time or route of administration, extent of drug effects in the user, or distinguish chronic from single dosing (Dawe et al., 2002; Kapur, 1993; Wolff et al., 1999). The biological half-life of lipophilic THC creates substantial individual variability in absorption, distribution, and elimination kinetics. Complete elimination of a single dose from urinary fuids might take more than 30 days (Adams & Martin, 1996; Buchan et al., 2002; Dawe et al., 2002). Repeated use results in accumulation of THC and its metabolites in the body, detectable up to 42 (and recorded up to 77) days after cessation of chronic, heavy intake (Vereby & Buchan, 1997; Wolff et al., 1999). A positive result could occur in a chronic heavy user who quit several weeks ago or a non-user with recent passive exposure to cannabis smoke (Hayden, 1991; Heustis & Cone, 1998; Wolff et al., 1999). However, although urinary cannabinoids cannot be used to reliably predict the recency of ingestion, semi-quantitative analysis can be used to monitor the fluctuations of urinary THC concentration over time, and serve to verify respondents’ self-reports (AACAP, 1997; Yacoubian, 2000).
Typically detecting recent (2-3 days) drug use, most biological measures provide no information on the severity of the underlying clinical syndrome or the cumulative effects of drug use (Carroll, 1995; Nathan, 1996; Saunders, Aasland, Amunsden, & Grant, 1993). Their low-to-moderate sensitivities and specificities for detecting drug dependants clearly limits their ability to identify at-risk and harmful drug use (Bohn, Babor, & Kranzler, 1995; Nathan, 1996; Poikolainen, 1999; Saunders et al., 1993). The mere presence of urinary or plasma cannabinoids does not necessarily indicate a pattern of dependent, harmful, or risky use (Carroll, 1995; Cone & Johnson, 1986; James &
Moore, 1999; Riley, Lu, & Thylor, 2000; Schwartz, 1988). Given their inability to provide information on the negative psychological, occupational, social and physical consequences of drug use, urine assay results should not be viewed as the definitive ‘gold’ standard to determine either recent or problematic cannabis use (Fals-Stewart, O’Farrell, Freitas, McFarlin, & Rutigliano, 2000; Heustis & Cone, 1998).
Major drawbacks of biological screens include high cost of laboratory equipment, expertise needed for results interpretation, and the invasiveness of some tests (Buchan et al., 2002; Fals-Stewart et al., 2000; WHOAWG, 2002). For urine, standard laboratory procedure involves an initial qualitative (positive/negative) screen test followed by quantitative confirmation. Commonly used qualitative screens (e.g., EMIT, homogeneous enzyme immunoassay) are generally reliable and valid, but unless confirmed with quantitative analysis, can yield false-positive results (Wolff et al., 1999). The most sophisticated ‘gold’ standard confirmatory technique available for routine urine screening is gas chromatography coupled to mass spectrometry (GC-MS). GC-MS is, however, an expensive procedure. Results are usually not available for at least 48 hours, which is an obvious barrier for rapid primary health care screening (Buchan et al., 2002; Saunders & Aasland, 1987). On-site use of self-contained cannabis testing kits requires clinical interpretation of laboratory values, a highly specialized task (Wolff et al., 1999). Finally, the full potential of most innovative techniques (hair analysis, sweat, saliva and breath tests) for cannabis screening is not yet known (United Nations, 1998; Wolff et al., 1999) and their cost is currently prohibitive for routine use.
Urine currently remains the preferred, most reliable biological fluid for the routine analysis of illicit drugs and their metabolites (Wolff et al., 1999). Its many advantages include: ease of collection, little preparation required, ability to be monitored for adulteration; comparative non-invasiveness, acceptability, cost-efficiency; and corroborative utility for self-reports (AACAP, 1997; Buchan et al., 2002; Wolff et al., 1999). Despite proliferation and sophistication of biological tools for drug use screening, however, accurate interpretation of laboratory findings requires other contextual information, and urine toxicology is more appropriately used in an adjunct
role. To date, no laboratory test in itself provides adequate information for discriminating problematic from nonproblematic cannabis use (Fals-Stewart et al., 2000; Fleming, 2002; Wolff et al., 1999).
Collateral information
Collateral reports from a spouse, family member, friend or coworker are relatively inexpensive, flexible, less invasive, and usually better than biological measures for collecting information on continuous outcomes and historical information. Collateral reports do, however, present their own limitations (see Babor et al., 2002; Carey & Simons, 2000; Del Boca & Noll, 2000). These include: recruitment difficulties, and lack of collateral informants to provide reports; lack of independence of collateral report from the drug user’s own report; collaterals having limited opportunity to directly observe the drug-using behaviour; collaterals “punishing” respondents by reporting higher levels of consumption than they actually observe; collaterals being drug users themselves (issues of accuracy, cooperation, or collusion); and the inability of collateral to independently observe small behavioural changes over time. Ultimately, collateral data rarely provides any more information than that provided by drug users themselves (Carey & Simons, 2000).
Self-report approaches
Undoubtedly, self-report methods have become the dominant means for collecting drug use information (Babor & Del Boca, 1992; Dennis et al., 2000). As yet, no single questionnaire has been universally adopted by the drug field (McPherson & Hersch, 2000; Sobell, Kwan, & Sobell, 1995). The most common screening questions used by health professionals focus on quantity/frequency of drinking and drug use (Dawe et al., 2002).
(A) Quantity/Frequency (Q/F) scales
Consumption information is an indispensable component of any screen, and frequency of use “the best single indicator of drug use involvement” (Clark, Pollock, Mezzich, Cornelius, & Martin, 2001, p. 16). Individuals are typically asked, “On average, how many days per week do you drink/use drugs? On a typical day when you drink/use
drugs, how much do you use?” Quick and easy to administer, Q/F scales have much to commend them (Fleming, 2002). Compared to detailed calendar-type methods recording daily substance use over a targeted time interval, however, Q/F scales are less precise (Sobell & Sobell, 1992). Limitations in obtaining cannabis consumption information only include:
(a) no existing guidelines on, or even the ability to determine, what constitutes a robust ‘standard’ measure/unit of cannabis;
(b) no consensus about cannabis consumption levels above which intake is hazardous or risky. ‘Safe’ drug consumption levels or standards are subject to perennial debate;
(c) substantially different individual outcomes at similar consumption levels; (d) unknown threshold use level or pattern for development of dependence;
(e) an assumption of stable consumption patterns and no information on fluctuations and atypical periods, such as episodic “binge” use, which is often indicative of early stages of problem use;
(f) no indication of respondents’ concern about their own consumption;
(g) provides no information about drug-related medical and psychosocial problems; (h) unacceptably low sensitivities as screening procedures for low-level misuse; and (i) reliability and validity issues common to all self-reported health data (APA, 1994,
2000; Dawe et al., 2002; Maly, 1993; Saunders et al., 1993).
Measuring levels of drug use other than alcohol and tobacco is challenging (Carroll, 1995; Sobell et al., 1995). First, most drugs are illicit - a feature which could discourage honest reporting of their use. Second, since the purity (strength) of drugs varies considerably, the actual intake of illicit substances is extremely difficult to quantify (Sobell et al., 1995). THC concentration in cannabis varies substantially (Hall & Swift, 2000). The variability within common street units (such as the ‘tinny’, ‘bag’, ‘blunt’, ‘joint’, ‘cone’, ‘bullet’) is too high to render them useful for research or ituations requiring precision. These units may include varying mixes of tobacco. Further measurement problems derive from frequent sharing of cannabis (Hall & Swift, 2000). Third, route of administration (e.g., smoking, spotting, eating) affects THC delivery, thus speed of effect onset. There is no standard method of intake (Clark et al., 2001;
Hall & Swift, 2000). Besides considerable individual vulnerability to its effects, various other factors (purity, plant strain, cultivation, product used, amount, potency, administration route, titration, degree of intoxication sought and attained, and so on) determine THC bioavailability (Adams & Martin, 1996; Hall & Swift, 2000; Stephens et al., 2002).
Thus, in stark contrast to the marked dose-response curves and ethanol-proof data in alcohol measurement, quantification of historical cannabis use is far less amenable to standardisation/estimation (APA, 2000; Hall & Swift, 2000; Stephens et al., 2002). The assumption that ‘cannabis use = cannabis problems’ is patently erroneous. Currently, cannabis consumption patterns and levels unequivocally associated with negative consequences are unknown (Kandel & Chen, 2000). While there is consensus that “heavy” cannabis use can be harmful or risky, terms such as “experimental” “heavy” or “frequent” use differ across studies, or are often not clearly defined (Earlywine, 2002; Kandel & Chen, 2000; Matt & Wilson, 1994; Zimmer & Morgan, 1997a, 1997b). Frequent use, moreover, “is not a necessary condition for the development of dependence symptoms” (Dennis et al., 2002a, p. 6). Thus, while consumption information must always be included in drug screening, clinical efficiency and utility is enhanced by administration of brief standardised screening instruments in which self- reported consumption is just one of the measures. Such a screening procedure is less threatening for the cannabis user, increasing the likelihood of eliciting reliable, valid information about drug use and attendant problems (Dawe et al., 2002; Del Boca & Noll, 2000; Dennis et al., 2000; Fleming, 2002; NHC, 1999).
(B) Standardised screening questionnaires
Standardised, self-administered computerized or paper-and-pencil questionnaires and clinical or nonclinical structured interviews have decided advantages over laboratory tests and clinical examination procedures with regard to cost and time consumption, acceptability, intrusiveness, level of administration and interpretive skill required. These procedures are also flexible, sensitive, potentially more comprehensive and clinically useful (Babor & Higgins-Biddle, 2000; Carroll, 1995; Dawe et al., 2002; WHOAWG, 2002). Important features are a high response rate, accurate sampling, and automatic
elimination of requirement for training and inter-rater reliability studies. Interviewer bias, which can enter the lengthy unstructured interview situation, is removed or greatly reduced by administration of short, standardised questionnaires with scoring guides or templates requiring minimal interpretation (Carroll, 1995; Dawe et al., 2002).
Self-report methods, however, rely on respondent veracity. Invalid response tendencies, including ‘faking good’, ‘faking bad’, inattention, and careless or random responding may be due to intentional efforts by the subject to manipulate the impression they wish to make upon others, or may be associated with lack of insight, defensiveness, poor attention, cognitive or reading deficits (Del Boca & Noll, 2000; Winters et al., 1991). A brief review of these issues follows.
Reliability and validity of self-reported cannabis use
Reliability and validity of self-reports in drug research has evoked vigorous debate (see Babor & Del Boca, 1992; Del Boca & Noll, 2000; Dawe et al., 2002). The negative pole is exemplified thus: “Substance abuse characteristics are unmatched by those of most other diseases. Denial, deception, and distortion are its handmaidens” (Schwartz & Wirtz, 1990, p. 38). Contradicting such (extremist) claims, comprehensive reviews of the reliability and validity of self-reported drug use among adults (e.g., Maisto, McKay & Connors, 1990; Spooner & Flaherty, 1992) found these generally reliable and valid, with greater consistency found for cannabis than other drug classes (see Darke, 1998, for more comprehensive review). Self-reported adolescent cannabis use is also generally reliable (Dennis et al., 2002b; Gignac et al., 2005; Lynskey, Fergusson, & Horwood, 1998).
Nonetheless, self-reported drug use cannot be uncritically accepted. The obvious intent of the questions in brief, simple self-report instruments renders them vulnerable to deliberate falsification or denial (Nathan, 1996). Mixed messages from ongoing legal, medical, clinical and scientific controversy about potential cannabis-related harm confuse public perceptions of risks involved (Hall, 1999; Hall & Solowij, 1998). Cannabis users tend to view their use as harmless and congruent with their lifestyle, have little interest in quitting, or lack motivation to accurately recall their use patterns or
problems (Buchan et al., 2002; Frances, First, & Pincus, 1995). While immaturity or lack of insight can interfere with an adolescent’s ability or motivation to introspect objectively, normal adolescence attributes (e.g., anti-authoritarianism, defiance) can curb motivation to cooperate, especially when under coercion to disclose (Spooner et al., 1996). Adolescents are less likely to report cannabis use if their parents are present, but more likely if friends or peers are present (Reid et al., 2000).
Denial and other response biases can impact on drug use recall, especially when respondents perceive response-contingent repercussions (Del Boca & Noll, 2000; Finch & Strang, 1998; Nathan, 1996). Illicit drug users face strong disincentives to divulge drug use, especially under circumstances of legal entanglement (arrests, probation, parole, intoxicated driving) (Buchan et al., 2002; Morral, McCaffrey, & Iguchi, 2000; Spooner & Flaherty, 1992). Under-reporting cannabis use may not be confined to users (Struve et al., 2000). In this study, after 8 weeks of intensive assessment including bi- weekly urine screens, 20% current users and presumptive non-user controls had deliberately falsified reports of their cannabis use. Mieczkowski and associates (1998) found, however, that while arrestees’ urine and hair bioassays indicated significantly more drug use than predicated by self-reports…“if youth do admit to drug use, it is likely to be marijuana use as opposed to any other drug” (p. 1565).
Not surprisingly, validity of self-reported drug use is biased by the assessment context (a job application, jail, treatment entry, a research interview, the presence of peers or parents), interviewer characteristics, and perceived threat to confidentiality (Finch & Strang, 1998; Matt & Wilson, 1994). The bias direction, however, is not consistent by context. Contrary to popular belief, evidence suggests that individuals are more likely to over-report than under-report drug use (Dawe et al., 2002). In contrast to Struve (2000), over-reporting occurred in an AIDS outreach study where access to treatment was limited to users (Dennis et al., 2000). Several other studies found self-reported cannabis use generally valid, detecting more use than rigorous laboratory tests, on-site qualitative tests, and collateral reports (e.g., Babor et al., 2002; Buchan et al., 2002; Darke, 1998; Dennis et al., 2002b; Stephens et al., 2002; Swift et al., 1998b). Research specifically designed to evaluate this issue (Buchan et al., 2002) found disagreements went both
ways. When validated against the gas chromatography/mass spectrometry (GC/MS; ‘gold’ standard) technique, however, self-reported past month cannabis use among 12- 18 year-old adolescents was always higher than the GC/MS findings. Conclusions were that given (1) that both self-report and laboratory measures have their own unique sources of error, and (2) the many considerations in the metabolism and excretion of cannabis (window of detection), these findings highlight the advantages of collecting multiple sources of consumption data. False positives in laboratory tests are often an artifact of decomposing metabolites (Buchan et al., 2002; Dennis et al., 2002b). Similar results emerged from another adolescent study (Akinci, Tarter & Kirisci, 2001) in which 13% (n=200) inaccurately reported cannabis use, with most discrepancy explained by over-reporting. A recent adolescent study (n=207) found that 97% youth directly reporting no past month use in a structured interview obtained a negative urine, while 79% who endorsed cannabis abuse/dependence had a positive urine screen. Toxicology screens and parental reports were less sensitive and specific (Gignac et al., 2005).
A study specifically designed to assess the validity of self-reported cannabis among polydrug users (Martin, Wilkinson, & Kapur, 1988) provided “strong evidence of the validity of self-reported cannabis use” (p. 149). In appropriate conditions (assurance of confidentiality, no fear of repercussions from disclosing drug consumption) multiple drug users do give valid self-reports, at least for cannabis use (Martin et al., 1988). Self- reported cannabis use was independently verified by urinalysis and collateral reports in a clinical trial for treatment of cannabis dependence in the United States (Babor et al., 2002), among Australian adolescents by correlating laboratory urine reports with self- reports (Copeland et al., 2001a; Martin et al., 2005), and in observational research among Australian long-term cannabis users (Swift et al., 1998b, 2000). Overall concordance was very high across these diverse contexts. Most discordance was due to either participant over-reporting or inadequate sensitivity of the screening test, prompting the conclusion that cannabis users do not systematically underreport their cannabis use (Babor et al., 2002; Swift et al., 1998a, 2000). High agreement between baseline and 10-year retrospective accounts of cannabis use among a general population sample was reported (Shillington, Cottler, Mager, & Compton, 1995). Cannabis had the
highest agreement rates of all drugs (86% lifetime use and 99.5% age of onset). Given the 10-year interval, this “highly accurate recall history” for an illicit drug was “particularly remarkable” (p. 106). Finally, among a New Zealand clinical sample of cannabis users assured of confidentiality and anonymity, Bashford (2000) reported both pre- and post-treatment coefficients between self-reported cannabis use and laboratory verifiers. The close correspondence between these measures provided strong support for the validity of these polydrug users’ cannabis accounts.
This brief review clarifies that, rather than an either/or phenomenon, validity of self- reported cannabis use is context-dependent, varying with the data collection method and respondent characteristics. Factors impacting on reliability and accuracy of self-reports include:
(a) respondent characteristics such as intelligence, cognitive deficits, and motivation. Large doses of cannabis, for example, can impair cognitive abilities related to