Tobacco Institute
A Review of: Smoking-Related Deaths and Financial Costs (Of the Preliminary Draft Presented by the Office of Technology Assessment on May 10, 1985)
Fields
Annotations
- 1. Sterling, T. Author
- Affiliation:
Simon Fraser University
- Affiliation:
Document Images
A REVIEW OF: SMDKIIrj-REIAM DEATf1S ADID FINANCIAL CCQSTS
(Of the Preli.annazy Draft Prepared by the
Office of Technwlogy Assessment on May 10, 1985)
T. Sterling, Pfa. D.
A. Arundel, B.A.
-J. Weinkan, D. Sc.
Faculty of Applied Science and School of Canputer Science
Si.mon Fraser Uizi..vvessity
Btsnaby, British Colvmbi.a V5A 1S6
June, 1985
*Dr. Sterling can be reached by phone at (604) 291-4685
(604) 733-1348
TIMN 308164

ACKNOWLEDGEMENTS
We thank Mary Hehn, Ted Irwin, Wilf Rosenbaum and Mary
Weinkam for valuable assistance in programming and manuscript
preparation.
TIMN 308165

INTRODIICTION
The Office of Technology Assessment, at the request of
Congressman Fortney H..(Pete) Stark, prepared an analysis of the
"Financial Costs of Smoking" for the Subcommittee on Health of
the House Ways and Means Committee, with special attention on
the impact of smoking-related disease on the Medicare and
Medicaid program. A workshop was organized by Karl Kronebusch
and staff of the OTA on April 9 and resulted in a Preliminary
Draft: Smokinq-Related Deaths and Financial Costs, of May 10,
1985. (We will refer to this document henceforth as the OTA
Draft.) The OTA Draft estimates smoking-related deaths and the
number of life-years lost due to smoking-related disease
categories and discusses the health care costs and lost
productivity costs for each category. Also planned was a
discussion of how health care costs and costs of other social
programs might change as a result of reduced smoking.
Congressman Donald_Sundquist asked us to review the OTA
Draft and the assumptions on which its calculations were based.
In agreeing to do so, we intend to provide OTA staff with a
constructive critical evaluation of the strength, weaknesses,
and validity of the available database on which any estimates of
costs will be based in the final analysis. in order to help
reach a consensus we have asked a number of knowledgeable
colleagues to comment on our review. These reviews will be
passed on to OTA. We also intend to discuss our review with OTA
staff and consultants.
1
.
.~.~ 308166
TI

The OTA Draft bases estimates of attributable risk due to
smoking on specific sets of data and procedures. There are a
number of concerns about the suitability of the OTA methodology.
The most important are:
1. The original data and the analysis used in the OTA Draft
ignore the significant confounding between smoking and other
factors that are associated with disease, especially
hazardous occupations. The prevalence of smoking differs
substantially for different occupations and is highest for
those occupations where workers are exposed to toxic and
carcinogenic materials.
2. Employment opportunities vary by the health status of
individuals and so, of course, do mortality and years of
In
survival. Estimates of expected disease frequencies can only
be made from populations with similar employment patterns.
This effect has been called the 'healthy worker effect' for
cohorts of employed workers or the 'volunteer effect' for
cohorts of volunteers enrolled in studies on the effect of
various risk factors on mortality rates.
consequence, estimates of attributable risk due to tobacco
use are riot valid if these estimates are derived from the
comparison of populations that differ substantially with respect
to employment patterns and occupational hazards, especially the
possibility of exposure to toxic chemicals, dusts or fumes.
A background discussion in Section I will deal with these
two major variables that affect the attributable risks for
2
,TIMN 308167

smoking. Both of these variables were neglected in the OTA
Draft. Section II will evaluate the procedure used in the OTA
Draft to compute the attributable risks of smoking and Section
III will briefly review the frequency with which smokers and
nonsmokers require medical care. Our final discussion in Section
IV will deal with the assertj,o_n that smoking shortens the life
span of smokers. There ar fve appendices that provide
additional support to the discussion in the text.
SECTION I: CONSEQUENCES OF CONFOUNDING
1. Smokinq Characteristics ~j Ty e of Employment
Smokers are more likely to work in occupations with
exposure to toxic dusts and fumes than non-smokers. This
relationship, and the relationship between smoking and many
other environmental and cultural factors, was noted by Haenszel
(1956) almost 30 years ago. The basic confounding between
smoking and.occupation and other potential risk factors means
that many cancers that are attributed to smoking may in fact
have occupational or other etiologies.
The National Center for Health Statistics obtained
information on smoking and occupation for members of 39,011
households in its yearly Household Interview Survey (HIS). This
data was used by Sterling (1976; 1978a) to determine the
prevalence of smoking by occupation. Table 1 compares 40
occupational categories with the highest smoking prevalence and
3
,rIMN 308168

TABLE 1: Comparison of 40 Occupation Categories with the Highest
and Lowest Smoking Prevalence Among White Males
Prevalence Percent
Type of worker Highest Lowest Highest Lowest
40 40 40 40
Blue collar 29 4 72.5 10.0
Service 9 7 22.5 17.5
White collar,
Technical,
Professional 2 29 5.0 72.5
and managerial
Total No. of 40 40 100% 100%
Occupations
40 occupations with the lowest smoking prevalence for white
males. While 72.5% of occupations with the highest smoking
prevalence are blue-collar, that is, exposed to hazardous work
and toxic fumes and dusts with relatively high probability, only
10% of those with the lowest smoking prevalence fall into that
category. Conversely, only 5% of occupations with the highest
smoking prevalence but. 72.5% of occupations with the lowest
smoking prevalence are white-collar, technical, professional or
management occupations, that is, unlikely to be exposed to
hazards, fumes and dusts at 'the work place. In fact, detailed
inspection of occupations finds that those with high smoking
prevalence also tend to be dirty and hazardous while those with
low smoking prevalence tend to be clean and relatively free of
4
TIMN 308169

hazards.
The HIS 1970 data was also used to compare the occupation
of different series of smokers, nonsmokers, former smokers,
smokers of regular and filter cigarettes, smokers who started
early or late in life and of households with more or fewer
smoking members. Figure 1 summarizes the results for whites.
(The same observations hold for blacks.) The proportions along
each horizontal line represents the percentage of individuals
who are either blue-collar or have
professional/managerial/proprietary, or other jobs. For example,
for white males who smoked 20 or more cigarettes a day, 44.1%
are blue-collar, only 6.2% are professional/management, and the
rest are in other occupations. Of individuals who smoked between
10 and .19 cigarettes a day, 42.7% areblue-collar and 5.8%
professional/management.
For nonsmokers, 37.5% are blue-collar while 10.9% are
professional/managerial. As a consequence, any comparison of
those who smoke with those who smoke less also compares groups
with different proportions of individuals exposed to toxic .
dusts, fumes, and hazardous conditions. It is not possible to
determine the cause of a difference in mortality or morbidity
between groups that differ with respect to smoking, without
analysing the occupational distributions of the two groups. In
short, it is not clear if smoking kills workers or working kills
smokers.
5
, ViM.S 308170

TOTAL NUMBER
ShIOKING FOR ALL PERCENT
HABITS . OCCUPATIONS BLUE COLLAR
Current Amount Smoked
Smoke 20+ 8.951
Smoke 10-19 2,589
Smoke 1- 0 1,572
None 18,203
Smoking Ststus
Never Smoked 9,694
Former Smokers 0.609
Type o/ ClQerette
Smoke Rsyular 3,638
Smoke Fblter 8,778
Ape Surted
Younger thin 20 9,E$5
Older than 20 1.744
i
,
50..
5.6 49.2 '
9.9 143.2
10.8 .
11.0
E
f
Mothers of Intants -
Smok(n0 389 49.1 103 69.4
Not Smoktng 828 38.7 U.1 52.8
Houtehold Smokiny
Husbsnd & Wl/e 2,701
Hutband Only 2,965
Wi/e Only 1,610
Ndther 6.391
f
I
30.5
30.9
N.1
42.7
138.8
PERCENT PERCENT
PROFESSIONAL OTHER OCCUPATION
41.8
,
41.3
I
41.9
50.3
48.6
46.2
46.9 3.9
7.2
141.8
43.7
133.3
27.0
u0
46.1
40.9
t
49.0
47.4
45.0
139.8
Figure 1. Cocnparison of proportions of blue-collar and professional workers in series of white
males and within
households headed by white mal.es, differentiated by smoking habit. Comparisons based ori data
obtained during the
1970 Household Interview Survey of the U.S. National Center for Health Statistics. Each horizontal
line represents
the proportion of individuals.who are either blue-collar workers or professionals, managers, or
proprietors.
From Sterling, 1978b.
9
'I
~_A

Independent verification of our results comes from a study
by Friedman (1973) which found that smokers enrolled in the
Permanente Insurance Plan had occupations with much higher
exposure to industrial hazards, especially fumes and dusts, than
nonsmokers.
There is also independent verification that when smoking
and occupational effects are explored in occupational health
studies, the observed effect of smoking often is much diminished
or entirely lacking compared to the effect of occupation. In
fact there are a number of studies in which the relative risk
and sometimes the absolute risk for nonsmokers exceeds that for
smokers. (See for instance Axelson 1978, Dahlgren 1979, Pinto
1978, Wagoner 1975, Weiss 1976, 1980 and the discussion in
Sterling 1983.)
Occupational effects become even more complex and perhaps
more important when differences between blacks and whites and
the sexes are considered. Black lung cancer mortality rates are
significantly higher than the rates for whites, with a 1976
age-standardized rate per 100,000 of 93 for black males compared
to 63 for white males and 19.5 for black females compared to
17.4 for white females (NCI, 1978). The difference cannot be
attributed to smoking, as even though fewer blacks than whites
are exsmokers, more blacks are never smokers and black smokers
consume fewer cigarettes than white smokers (Sterling, 1978a).
The difference in mortality rates by race is probably partly a
result of differences in occupational exposures. Boucot (1970)
7
0
,VIM-S 308172

noted that almost twice as many blacks were employed in jobs
with an increased cancer risk compared to whites; 22.6% versus
13.5%. Similar results were found by Mancuso (1975).
Women are exposed to hazards at work-as well as in the
household. (We consider housework as an occupation.) Housewives
tend to smoke approximately as much as other women and, at the
same time, they are exposed to a variety of hazards, such as
solvents or combustion byproducts from gas stoves&hich ~
represen more than a trivial source of toxic and carcinogenic
materials ( terling 1979a; 1981). Also, as shown in Figure 1,
there is a strong association between the number of women and
men smokers in the household 4nd the occupation of the husband.
A large number of women who smoke also.have a blue-collar family
background and may be exposed not only to hazards at work on
their own but also to toxic materials brought home from work on
the hair, skin and clothing of their spouses.
The distribution of smoking by occupation (and consequently
also by socioeconomic class) indicates that the occupation of
the smoker is one of the'variables that must be included in any,
evaluation of attributable risks. All cancers presently
associated with smoking are also associated with occupation,
such as cancer of the esophagus, bladder, pancreas and oral
cavity, but especially of the lung. Yet, the effect of
occupation is not considered in most of the studies on which the
OTA Draft report is based. The OTA's own estimates for the
attributable risk for cancer and heart disease did not consider
8
t
TIMN 308173
