Evidence for Genetic Influence on Tobacco Use in Humans 159subsequent twin studies have supported the presence of sepa-
rate but correlated sources of genetic and environmental
sources of variation in smoking initiation and persistence
and/or dependence (Koopmans et al., 1997; True et al.,
1997). Of these papers, only Kendler et al. (1999) utilized a
direct measure of dependence, a factor derived from the FTQ,
a class II phenotype according to our classi“cation scheme.
(See following discussion of the genetics of dependence.)
The appreciation of smoking behavior as multidimen-
sional in behavior genetic studies is relatively recent and
parallels similar work being conducted in the areas of alcohol
and substance use. Along with the genetic models applied by
Heath and Martin (1993) to smoking initiation and persis-
tence, Koopmans et al. (1997) also examined the goodness-
of-“t of a combined liability model “rst developed by Heath,
Meyer, Jardine, and Martin (1991) and found it to provide the
best “t to adolescent twin covariance for smoking initiation
and amount smoked. Their analysis provided results similar
to that found by Kendler et al. (1999): separate genetic liabil-
ities for initiation and quantity were present, accounting for
39% of initiation and 86% in amount smoked. Heath, Kirk,
Meyer, and Martin (1999) also provide evidence that dif-
ferent genetic and environmental factors are present for risk
of initiation (11%...74% genetic) and for the age at onset
(50%...60% genetic) of smoking, with shared environmental
effects playing a larger role in initiation than in the age at
onset of smoking. Madden et al. (1999) provide further evi-
dence of the dimensionality of smoking behavior by using a
correlated liability dimensions model to determine that less
than 40% of the total genetic variance in smoking persistence
was accounted for by the same genetic factors that increased
risk of smoking initiation, a percentage that decreased with
the age of the twins under study. It is important to remember
that these investigations based their “ndings on dif fering
combinations of class I phenotypes determined from survey-
type questionnaires.
In a study of direct relevance to the present chapter,
Stallings, Hewitt, Beresford, Heath, and Eaves (1999) exam-
ined the contribution of genetic and environmental in”uences
to age at onset and, for the “rst time of which we are aware,
thetimein years between “rst use and regular use of tobacco
and alcohol. The latency in progression to regular use quali-
“es as a class III phenotype according to our system, and
represents an advance in the use of more descriptive tobacco-
related phenotypes. These investigators found little evidence
for genetic involvement in several milestones (age at “rst
use, age at daily use of at least one cigarette per day, age at
daily use of at least 10 cigarettes per day, all class I pheno-
types). In contrast, 37% of the variation in the speed with
which twins progressed from “rst use to regular smoking of
at least 10 cigarettes per day could be attributed to genetic
factors.
With regard to existing molecular genetic literature involv-
ing smoking-related phenotypes, most reported studies exam-
ined class I, and occasionally class II, phenotypes measured
retrospectively (Bierut et al., 2000; Boustead, Taber, Idle, &
Cholerton, 1997; Cholerton, Boustead, Taber, Arpanahi, &
Idle, 1996; Comings et al., 1996; Lerman et al., 1998, 1999;
Noble et al., 1994; Pianezza, Sellers, & Tyndale, 1998; Sabol
et al., 1999; Shields et al., 1998). For example, Spitz et al.
(1998) report an association between the dopamine d2 recep-
tor, DRD2, alleles and likelihood of being a smoker based on
assessments of self-reported smoking behavior that was as-
sessed years, or even decades, after the onset of smoking.
Nicotine dependence also was assessed retrospectively;
as such, subjects (former smokers) were asked to recall
speci“cs about their smoking behavior, in an attempt to an-
swer items from the FTQ. Retrospective case-control designs
are subject to limitations of recall bias. Many of these studies
have not yet received independent con“rmation.
While the existing behavioral genetic and genetic epi-
demiologic literature has just begun to scratch the surface
of the developmental aspect of smoking behavior by using
class III phenotypes (Stallings et al., 1999), the few studies
that have been reported suggest that not only do different de-
velopmental phases of smoking have identi“able genetic
sources of variance (Carmelli, Swan, Robinette, & Fabsitz,
1992; Heath & Martin, 1993; Swan, Reed, & Carmelli, 2000)
but that the underlying genetic factors involved in each may
be different (Heath & Martin, 1993; Kendler et al., 1999;
Koopmans et al., 1997; Madden et al., 1999; Swan et al.,
2000; True et al., 1997). Therefore, it is within the realm of
possibility that molecular genetic investigations of adult
smokers that identify associations between the likelihood of
being a smoker (as assessed by single-point measures) while
perhaps being of relevance to understanding factors involved
in the maintenance of smoking, are less relevantto the un-
derstanding of genetic factors in the acquisition, cessation,
and relapse phases of smoking.Tobacco Dependence: A Construct in Need
of RefinementThe de“nition and measurement of tobacco dependence
continues to represent a challenge to the “eld. While it
was pointed out early (Hughes, 1985; also Lombardo,
Hughes, & Fross, 1988) that dependence has several dimen-
sions including physical, behavioral, and psychological
components, the assessment of tobacco dependence has
relied largely on the Fagerström Tolerance Questionnaire