Segregation analyses of lung- and smoking-associated cancers

Given the evidence for familial aggregation of lung
and other smoking-associated cancers, after ac-
counting for personal tobacco use and occupa-
tional/industrial risk factors, segregation analyses
have been performed to determine whether pat-
terns of transmission consistent with at least one
major, high-penetrance genetic locus may be in-
volved in lung cancer risk.
Sellers et al. [80] performed genetic segregation
analyses on the lung cancer proband families of Ooi
et al. [54] described above. The trait was expressed
as a dichotomy, affected or unaffected with lung
cancer. The analyses used the general transmission
probability model [81], which allows for variable
age of onset of the lung cancer [82–84]. The likeli-
hood of themodelswas calculated using a correction
factor appropriate for single ascertainment [85,86],
i.e., conditioning the likelihood of each pedigree on
the probands being affected by their ages at exami-
nation or death.
Age of onset of lung cancer was assumed to fol-
low a logistic distribution that depended on pack-
years of cigarette consumption and its square, an
age coefficient and a baseline parameter. Results in-
dicated compatibility of the data with Mendelian
codominant inheritance of a rare major autosomal
gene that produces earlier age of onset of the can-
cer. Segregation at this putative locus could account
for 69 and 47%of the cumulative incidence of lung
cancer in individuals up to ages 50 and 60, respec-
tively. The genewas predicted to be involved in only
22% of all lung cancers in persons up to age 70, a
reflection of an increasing proportion of noncarriers
succumbing to the effects of long-term exposure to
tobacco [80,87].
Additional segregation analysis of these families
was performed to determine whether evidence ex-
ists for a major gene that increases susceptibility to
a group of smoking-associated cancers rather than
just lung cancer alone. The trait was defined as
unaffected or affected with cancer at any of the
following sites: lung, lip, oral cavity, esophagus, na-
sopharynx, trachea, bronchus, larynx, cervix, blad-
der, kidney, colon/rectum, and pancreas. The results
were compatible with segregation of a major gene
that influences age-of-onset of cancer. The hypothe-
ses ofMendelian dominant, codominant, and reces-
sive inheritance could not be rejected but, according
to Akaike’s Information Criterion [88], Mendelian
codominant inheritance provided the best fit to
the data [89]. Additional analyses suggest that bet-
ter fit of Mendelian inheritance of an allele that
acts with smoking to influence the risk of can-
cer is obtained if a somewhat different cluster of
smoking-associated cancers is considered “affected”:
lung, oral cavity, esophagus, nasopharynx, lar-
ynx, pancreas, bladder, kidney, and uterine cervix
[90].
Segregation analyses of these data ([91] and
our unpublished results) using Class A regressive
models showed significant evidence for a poly-
genic/multifactorial component in addition to the
major gene component described above. Inclusion
of this polygenic/multifactorial component signifi-
cantly improved the fit of the model to the data
without changing the basic conclusions of the pre-
vious analyses, i.e., that evidence exists for a major
locus with a codominant susceptibility allele that
acts in conjunction with smoking, and that all mod-
els excluding such amajor gene effectwere rejected.
Gauderman et al. [92] reanalyzed these same data
using a Gibbs sampler method to examine gene by
environment interactions and found evidence for a
major dominant susceptibility locus that acts in con-
junctionwith cigarette smoking to increase risk; this
model was very similar to the previous results since
the codominant Mendelian models predicted very
small numbers of homozygous susceptibility allele
carriers.
Yang and her coworkers performed complex seg-
regation analysis on the families of nonsmoking
lung cancer probands in metropolitan Detroit [93].
Evidence was found for Mendelian codominant in-
heritance with modifying effects of smoking and
chronic bronchitis in families of nonsmoking cases
diagnosed at ages 40–59. The estimated risk allele
frequency was 0.004. While homozygous individu-
als with the risk allele are rare in the study popula-
tion, penetrance was very high for early-onset lung
cancer (85% in males and 74% in females by age
60). The probability of developing lung cancer by
age 60 in individuals heterozygous for the rare al-
lele was low in the absence of smoking and chronic
bronchitis (7% in males and 4% in females) but in
the presence of these risk factors it increased to 85%
in males and 74% in females, which was the same
level predicted for homozygotes. The attributable
risk associatedwith the high-risk allele declineswith
age, when the role of tobacco smoking and chronic
bronchitis become more important.
Wu et al. [94] performed complex segregation
analysis of families of 125 female, nonsmoking
lung cancer probands in Taiwan. These lung cancer
probandswere diagnosedwith lung cancer between
1992 and 2002 at two hospitals in Taiwan. Complete
data on patients, spouses and first-degree relatives
were collected for 108 families.Data collected on the
patients and their relatives included demographic,
life-style, and medical history variables. Complex
segregation analysis using logistic models for age at
onset, including pack-years of cigarette smoking in
the model was performed on 58 of these families.
An ascertainment correction was made using the
phenotype of the probands, but this may have been
inadequate since the 58 families were a subset of
the 108 families where there was at least one addi-
tional affected relative in the family. TheMendelian
codominant model that included risk due to per-
sonal smoking fit the data best, significantly better
than the sporadic or purely environmental mod-
els. This model was not rejected against the general
model in an early-onset (less than 60 years) subset
of the families but was rejected in the later-onset
families and the total dataset.
Taken together, the Taiwan, Detroit, and
Louisiana studies share remarkably similar results
and demonstrate statistical evidence for at least one
major gene that acts in conjunction with personal
smoking and possibly chronic bronchitis to increase
risk of lung cancer.