J Clin Periodontol 2003; 30: 107–113 Copyright r Blackwell Munksgaard 2003

Printed in Denmark. All rights reserved

Tobacco smoking and risk for Jan Bergström

Institute of Odontology, Karolinska Institutet,
Stockholm, Sweden

periodontal disease
Bergström J: Tobacco smoking and risk for periodontal disease. J Clin Periodontol
Blackwell 2003; 30: 107–113. r Blackwell Munksgaard, 2003.

Abstract
Objective: The magnitude of risk for periodontal disease associated with smoking
was investigated by exploring the interrelationships between definition of disease,
prevalence and relative risk in a population of dentally aware individuals.
Materials and Methods: A total of 133 smokers and 242 non-smokers in the age
range 20–69 years served as the database. Criteria based on clinical probing of pocket
depth and radiographic measurement of bone height were used for the purpose of
disease definition. Various pocket frequency cutoff points for two probing depth
levels, X5 and X6 mm, and, in addition, the 30th, 25th, and 20th percentiles of the
bone height distribution were alternately selected as criteria.
Results: Disease definition had an impact on the prevalence, and the relative risk
varied as a function of the prevalence. For a broad definition of disease such as 1% of
pockets X5 mm, the prevalence for smokers was approximately twice the prevalence
for non-smokers and the odds ratio (OR) was 3.0. A narrower definition such as 15%
of pockets X5 mm resulted in decreased prevalences for both smokers and non-
smokers and an OR of 12.1. Increasing exposures correlated with greater magnitudes
of risk. Heavy exposure was associated with greater risk than light exposure. For the
combination of a narrow disease definition and heavy exposure, the relative risk was
considerable (OR 9.8–20.3).
Conclusion: Smoking-associated relative risk is dependent on definition of disease Key words: periodontitis; relative risk; risk
and prevalence. Given other factors, a narrow definition will result in a low prevalence assessment; smoking; tobacco
and a high risk and, vice versa, for a broader definition, prevalence will be high and
the risk moderate. Accepted for publication 2 July 2001

The risk of attracting periodontal dis-
ease in association with tobacco smok-
ing has been estimated in several studies
over the years. The variation in magni-
tude of the risk estimates reported,
however, is considerable. Risk estimates
found from cross-sectional and case–
control studies vary between 1.4 and
11.8 (Markkanen et al. 1985, Preber &
Bergström 1986, Bergström 1989, Ha-
ber & Kent 1992, Horning et al. 1992,
Haber et al. 1993, Hansen et al. 1993,
Locker & Leake 1993, Stoltenberg et al.
1993, Grossi et al. 1994, 1995, Linden
& Mullally 1994, Mullally & Linden
1996, Dolan et al. 1997, Norderyd &
Hugoson 1998, Moore et al. 1999, Paidi
et al. 1999, Wakai et al. 1999) and
those found from the few cohort
studies available vary between 4.6
and 6.3 (Ismail et al. 1990, Holm
1994, Norderyd et al. 1999, Skaleric &
Kovac-Kavcic 2000). Some reasons
for the large variation may be found
in the type of study and character and
size of the population studied. An
additional factor may be the definition
of disease. The wide variation among
studies regarding diagnostic criteria for
the determination and classification of
disease is striking. Among commonly
used determinants are, e.g., probing
depth, attachment level, bone height,
furcation involvement, tooth loss, as
well as composite indices. There is,
furthermore, a wide variation regarding
the choice of boundary or cutoff border-
line for the discrimination of disease
from nondisease, e.g., one pocket of
5 mm probing depth; one tooth with an
attachment loss of 7 mm; a majority of
teeth exhibiting alveolar bone loss
exceeding one-third of the root length,
etc. Apparently, therefore, the definition
of the disease is an important factor that
can be expected to contribute to the
variation in magnitude of the risk
estimate. We have previously reported
on the influence of smoking on the
periodontal health in subjects with
regular dental care habits (Bergström
& Eliasson 1987a, b, Bergström et al
2000a, b). To further explore the impact
of smoking on the periodontal health,
the purpose of the present study was to
investigate the relations between defini-
tion of disease, prevalence and relative
risk associated with smoking in this
population.
Materials and Methods
Study population
Two samples drawn from the same
population with a 10-year interval were
used as the database of the present
investigation. The population is char-
acterized by a high standard of dental
awareness as seen from, e.g., regular
108 Bergström
dental care habits, and a high standard
of oral health as seen from, e.g., a great
number of retained teeth. A detailed
description of the characteristics of the
population has been reported elsewhere
(Bergström & Eliasson 1985, Bergström
et al. 2000a). The first sample included
83 smokers and 110 non-smokers in the
age range 20–69 years (median 41
years, IR 31–50 years), and the second
sample included 50 smokers and 132
non-smokers in the same age range
(median 38 years, IR 31–51 years). In
addition, to increase statistical power
and estimate precision, analyses were
also performed for both samples
combined. The latter analyses included
63 individuals (17 smokers and 46 non-
smokers) who were part of both sam-
ples. All smokers were current smokers,
since for the purpose of the present
study former smokers were excluded.
The mean cigarette consumption and
smoking duration of smokers in sample
1 was 14.0 (95% CI 12.4–15.5) cig/day
and 21.7 (95% CI 19.4–24.0) years,
respectively. The corresponding means
in sample 2 were 13.3 (95% CI 11.2–
15.4) cig/day and 20.4 (95% CI 16.7–
24.0) years, respectively. The study
samples, referred to as cohorts
1 and 2, respectively, according to
gender and smoking are presented in
Table 1. The mean (SD) number of
retained teeth in smokers and non-
smokers was 27.0 (3.8) and 28.0 (2.8),
respectively, in cohort 1, and 28.0 (2.5)
and 28.7 (2.6), respectively, in cohort 2.
Although slightly lower in smokers,
these means were not statistically sig-
nificantly different when controlling
for age. The study was approved by
the local ethical committee of the
Karolinska Institute.
Clinical registration and classification
Several criteria based on clinical prob-
ing of pocket depth and radiographic
measurement of bone height were
alternately used for the purpose of
disease definition and case identifica-
tion. With regard to pocket probing
depth two levels (critical levels), 5 and
Table 1. Study cohorts; frequency distributions by gender and smoking
Cohort 1
men women
smoker 72 11
non-smoker 86 24
total 158 35
6 mm, were used. For a critical level of
5 mm, i.e., a probing depth of 5 mm or
greater, a pocket frequency of 1%, 5%,
10%, and 15%, respectively, was se-
lected as a cutoff point. For a critical
level of 6 mm, i.e., a probing depth of
6 mm or greater, pocket frequencies of
1%, 5%, and 10% were the selected
cutoff points. With regard to bone
height, the 30th, 25th, and 20th percen-
tiles of the distribution were used as
cutoff points. These cutoff points corre-
sponded to a mean bone height level of
82%, 80%, and 76%, respectively, in
cohort 1, and 83%, 82%, and 80%,
respectively, in cohort 2.
Probing depth was assessed at four
positions of all teeth, and bone height
was measured as a percentage of the
root length mesially and distally to all
teeth and the average across all single
measurements formed the bone height
level of the individual (Eliasson &
Bergström 1986).
The smoking exposure was expressed
in terms of consumption (cigarettes per
day), duration (number of years of
smoking), and lifetime exposure, i.e.,
the product of consumption and dura-
tion (cigarette-years). Analyses regard-
ing exposure were performed for both
cohorts combined only, with exposure
variables stratified into three subgroups.
The subgroups regarding consumption
were 1–9 cig/day (n 5 28, mean 5 3.8
cig/day), 10–19 cig/day (n 5 57,
mean 5 12.1 cig/day), 20 or more cig/
day (n 5 47, mean 5 21.5 cig/day); re-
garding duration 1–15 years (n 5 39,
mean 5 7.9 years), 16–30 years (n 5 53,
mean 5 20.4 years), more than 30 years
(n 5 41, mean 34.9 years); regarding
lifetime exposure 1–170 cigarette-years
(n 5 46, mean 5 74.5 cigarette-years),
171–350 cigarette-years (n 5 44,
mean 5 264.4 cigarette-years), more
than 350 cigarette-years (n 5 42,
mean 5 614.3 cigarette-years).
Statistics
The relative risk was estimated from the
odds ratio (OR) following logistic
regression with pocket frequency, bone
Cohort 2 Cohorts combined Total (%)
men women men women
40 10 112 21 133 (35)
92 40 178 64 242 (65)
132 50 290 85 375 (100)
height or combinations of pocket fre-
quency and bone height, respectively, as
the dependent dichotomous variable.
Pocket frequency was dichotomized
according to o1% (0) and X1% (1);
o5% (0) and X5% (1); o10% (0) and
X10% (1), etc. These steps were
labelled 1%, 5%, 10%, etc., cutoff
points. Following the percentile distri-
butions of the cohorts, bone height level
was dichotomized according to 430th
percentile (0) and 430th percentile (1);
425th percentile (0) and 425th per-
centile (1); 420th percentile (0) and
420th percentile (1). These bone height
levels were labelled 30th percentile,
25th percentile, and 20th percentile
cutoff points, respectively. The possible
confounding of other factors such as
age, plaque, calculus, gingival bleeding,
and number of teeth was tested in
preliminary analyses. It was revealed
that age exerted influence on the
dependent variables, whereas the effects
of the others were marginal, if any. Age,
therefore, was regarded as a confounder
and included in all regression analyses.
If not otherwise explained, therefore, all
results are valid after adjustment for
age. Statistical significance was ac-
cepted at po0.05.
Results
The prevalence according to pocket
frequency cutoff point for a critical
level of 5 mm together with the age-
adjusted smoking-associated ORs is
presented in Table 2. Regarding cohort
1, a 1% cutoff resulted in a prevalence
of 70% for smokers and 41% for
non-smokers and an OR of 2.5 (95% CI
1.3–5.0). At a 5% cutoff the prevalence
decreased to 41% for smokers and 15%
for non-smokers and the OR increased to
3.1 (95% CI 1.4–6.6). At 10% and 15%
cutoffs, the prevalence further de-
creased to 29% and 20%, respectively,
for smokers and 6% and 4%, respec-
tively, for non-smokers, the ORs being
4.9 (95% CI 1.9–12.9) and 5.5 (95% CI
1.7–17.7), respectively. The overall
prevalence related to these cutoff points
decreased from 53% to 11%. The risk
estimates were throughout statistically
significant (po0.01). A similar trend of
decreasing prevalence and increasing
risk with increasing cutoff point was
observed for cohort 2. As the cutoff
was shifted from 1% to 10%, the
prevalence decreased from 30% to
12% for smokers and from 16% to
2% for non-smokers, whereas the OR
 Smoking-associated risk 109

Table 2. Prevalence (%) and odds ratio (OR) according to pocket frequency cutoff point and cohort
Power Cohort 1 Cohort 2 Cohorts combined
frequency
cutoff (%) prevalence OR 95% CI p prevalence OR 95% CI p prevalence OR 95% CI p

S NS total S NS total S NS total

1 70 41 53 2.5 1.3–5.0 0.009 30 16 20 2.3 1.1–5.5 0.049 55 27 37 3.0 1.8–4.6 o0.001

5 41 15 26 3.1 1.4–6.6 0.003 16 04 07 5.8 1.6–21.4 0.008 30 08 16 4.7 2.5–9.0 o0.001
10 29 06 16 4.9 1.9–12.9 0.001 12 02 04 9.6 1.7–52.8 0.009 23 04 10 7.3 3.2–15.5 o0.001
15 20 04 11 5.5 1.7–17.7 0.004 08 00 02 — — — 17 02 07 12.1 3.9–37.1 o0.001
Critical probing depth level 5 mm. Adjusted for age. S 5 smoker, N 5 non-smoker.

Table 3. Prevalence (%) and odds ratio (OR) according to pocket frequency cutoff point and cohort
Power Cohort 1 Cohort 2 Cohorts combined
frequency
cutoff (%) prevalence OR 95% CI p prevalence OR 95% CI p prevalence OR 95% CI p

S NS total S NS total S NS total

1 51 27 37 2.1 1.1–4.0 0.033 20 08 12 3.0 1.1–8.6 0.036 39 17 25 2.9 1.7–4.9 o0.001

5 26 06 15 4.3 1.6–11.3 0.003 12 02 05 6.6 1.4–30.4 0.015 19 04 09 5.6 2.4–13.0 o0.001
10 17 03 09 5.7 2.0–21.5 0.010 10 00 03 — — — 14 01 06 12.7 3.5–45.0 o0.001
Critical probing depth level 6 mm. Adjusted for age. S 5 smoker, N 5 non-smoker.

Table 4. Prevalence (%) and odds ratio (OR) according to bone height cutoff point and cohort
Bone Cohort 1 Cohort 2 Cohorts combined
height
cutoff (%) prevalence OR 95% CI p prevalence OR 95% CI p prevalence OR 95% CI p

S NS total S NS total S NS total

30th percentile 49 15 30 4.2 1.8–9.6 0.001 45 26 31 2.8 1.1–7.0 0.031 48 21 30 4.3 2.3–8.2 o0.001
25th percentile 45 13 27 5.1 1.9–13.9 o0.001 40 19 25 4.2 1.5–11.8 0.005 45 14 25 7.5 3.7–15.3 o0.001
20th percentile 32 09 19 6.0 2.4–14.8 0.001 34 11 17 9.3 2.8–31.5 o0.001 35 09 18 8.1 3.8–17.5 o0.001
Adjusted for age. S 5 smoker, N 5 non-smoker.

increased from 2.3 (95% CI 1.0–5.5)
to 9.6 (95% CI 1.7–52.8). At a 15%
cutoff, no disease was present in non-
smokers and the OR could not be
estimated. The overall prevalence
decreased from 20% to 2% with
increasing cutoff point. The risk esti-
mates were throughout statistically sig-
nificant (po0.05 to po0.01), but
confidence intervals around those re-
lated to the greater cutoff points were
wide. The observed pattern became
even more evident for both cohorts
combined. The smoking-associated
OR increased from 3.0 (95% CI 1.8–
4.6) at a 1% cutoff to 12.1 (95% CI 3.9–
37.1) at a 15% cutoff in parallel with
decreasing prevalence rates. The risk
estimates were throughout highly sig-
nificant (po0.001).
The prevalence and smoking-asso-
ciated ORs according to pocket fre-
quency cutoff point for a critical level of
6 mm are presented in Table 3. Regard-
ing cohort 1, the OR increased from 2.1
(95% CI 1.1–4.0) to 5.7 (95% CI 2.0–
21.5) at 1% through 10% cutoff.
Regarding cohort 2 the OR increased
from 3.0 (95% CI 1.1–8.6) to 6.6 (95%
CI 1.4–30.4) as the cutoff was shifted
from 1% to 5%. At a 10% cutoff, the
OR could not be estimated since no
disease was present in non-smokers. For
both cohorts combined, the OR at 1%,
5%, and 10% cutoffs gradually in-
creased from 2.9 (95% CI 1.7–4.9) to
12.7 (95% CI 3.5–45.0) along with
decreasing prevalence rates in both
smokers and non-smokers. The risk
estimates were throughout statistically
significant (po0.05 to o0.001).
The prevalence and smoking-asso-
ciated ORs according to bone height
criteria are presented in Table 4.
Regarding cohort 1, the prevalence at
the 30th percentile bone height cutoff
was 49% and 15% for smokers and non-
smokers, respectively, and the OR was
4.2 (95% CI 1.8–9.6). At the 25th and
20th percentile cutoffs, the prevalence
dropped for both smokers and non-
smokers, whereas the OR increased to
5.1 (95% CI 1.9–13.9) and 6.0 (95% CI
2.4–14.8), respectively. Similar obser-
vations held true for cohort 2. For both
cohorts combined, the ORs related to
these percentile cutoffs were 4.3 (95%
CI 2.3–8.2), 7.5 (95% CI 3.7–15.3), and
8.1 (95% CI 3.8–17.5), respectively.
The risk estimates were throughout
statistically significant (po0.05 to
o0.001).
The age-adjusted smoking-associated
OR according to combinations of bone
height and 5 mm critical level pocket
frequency cutoff points is demonstrated
in Table 5 for both cohorts combined.
At the 30th percentile cutoff in combi-
nation with gradually increasing pocket
110 Bergström

frequency cutoffs, the OR increased combined. At each frequency cutoff, the

o0.001
o0.001
o0.001
from 5.4 (95% CI 2.8–10.4) to 8.3 ORs associated with consumption levels

p
(95% CI 3.3–20.8) as the prevalence 1–9 cig/day, 10–19 cig/day, and more
decreased from 39% to 21% for smo- than 19 cig/day, respectively, gradually
kers and from 12% to 3% for non- increased from approximately 2 to a

20.8
36.5
48.1
smokers. At the 25th percentile bone maximum of 10.7 (95% CI 4.1–28.1)
CI
10% pocket frequency cutoff

height cutoff in combination with the (Fig. 1a). The ORs associated with
same pocket frequency cutoffs, the OR smoking duration levels 1–15 years,
increased from 6.8 (95% CI 3.7–15.3) to 16–30 years, and more than 30 years
95%

3.3
3.5
4.8 11.4 (95% CI 3.5–36.5), and at the 20th gradually increased from 1.2 (95% CI
percentile bone height cutoff it in- 0.6–2.9) to 5.3 (95% CI 2.1–13.2) at a
creased from 10.7 (95% CI 4.7–24.7) 1% cutoff, from 2.3 (95% CI 0.7–7.8) to
8.3
11.4
15.3

to 15.3 (95% CI 4.8–48.1). The pre- 7.9 (95% CI 3.5–17.9) at a 5% cutoff,

OR

valence rates decreased from 35% to and from 4.2 (95% CI 0.9–19.0) to 9.8
20%, and from 31% to 20%, respec- (95% CI 3.7–26.0) at a 10% cutoff (Fig.
NS total

09
09
08

tively, for smokers and from 8% to 3%, 1b). Similarly, a gradual elevation of the
prevalence

and from 6% to 2%, respectively, for OR was observed along with lifetime
03
03
02

non-smokers. All estimates were highly exposure levels 1–170 cigarette-years,

significant (po0.001). 170–350 cigarette-years, and more than
21
20
20
S

350 cigarette-years from 2.2 (95% CI

Table 5. Prevalence (%) odds ratio (OR), and 95% confidence interval (CI) according to bone height and pocket frequency cutoff points

Exposure-related risk
1.1–4.4) at a 1% cutoff and light
o0.001
o0.001
o0.001

exposure (o170 cigarette-years) to

p

The age-adjusted ORs associated with 11.4 (95% CI 4.5–29.0) at a 10% cutoff
gradients of cigarette consumption, and heavy exposure (4350 cigarette-
smoking duration, and lifetime exposure years) (Fig. 1c). The risk estimates
16.6
27.2
39.6

according to pocket frequency cutoff associated with light exposure levels

5% pocket frequency cutoff

CI

point for a critical level of 5 mm are were generally rather weak and not
illustrated in Figs. 1a–c for both cohorts always statistically significant, whereas
95%

3.3
4.4
4.0
Critical probing depth level 5 mm. Both cohorts combined. Adjusted for age. S 5 smoker, NS 5 non-smoker.
7.4
10.9
12.5
OR

NS total

13
11
10
prevalence

05
03
02
27
25
25
S

o0.001
o0.001
o0.001
p
1% pocket frequency cutoff

10.4
15.3
24.7
CI
95%

2.8
3.7
4.7
15 10.7
21 5.4
18 6.8
OR

NS total
prevalence

12
08
06
39
35
31
S

30th percentile
25th percentile
20th percentile
Bone height

Fig 1. Odds ratios (95% CI) according to smoking exposure and pocket frequency cutoff
point. Critical level of probing depth: 5 mm. Cutoff points: 1%, 5%, and 10%. Both cohorts
cutoff

combined. (a) Cigarette consumption (cig/day). (b) Smoking duration (years). (c) Lifetime
exposure (cigarette-years).

estimates for moderate and heavy ex-
posures were highly significant, although
confidence intervals were wide. Similar
results were obtained using 6 mm prob-
ing depth or bone height as a basis for
disease definition (data not shown).
The relations between the smoking
exposure-associated relative risk and
combinations of bone height critical
levels and 5 mm pocket frequency cut-
off points are illustrated in Figs. 2a–c
regarding smoking duration and both
cohorts combined. For each level of
bone height, the OR gradually increased
with increases in exposure and pocket
frequency cutoff. The risk estimates
were normally highly significant, but
as seen in the figures the confidence
interval became wider as the OR
increased. The highest risk was ob-
served for the combination of heavy
exposure and great cutoff levels. The
risk run by a heavy smoker (more than
30 years’ duration) of having severe
disease (less than 76% bone height
Fig 2. Odds ratios (95% CI) according to smoking duration, bone height percentile cutoff
point and pocket frequency cutoff point. Critical level of probing depth: 5 mm. Cutoff points
of pocket frequency: 1%, 5%, and 10%. Both cohorts combined. (a) Bone height cutoff point:
30th percentile. (b) Bone height cutoff point: 25th percentile. (c) Bone height cutoff point:
20th percentile.
together with more than 10% of 5 mm
or deeper pockets) was approximately
20-fold that of a nonsmoker.
Discussion
Under the presumption that smoking
constitutes a risk for periodontal dis-
ease, the objective of the present
investigation was to explore the rela-
tions between the smoking-associated
relative risk, disease definition, and
prevalence of the disease. In the main,
the observations showed that the esti-
mated risk was moderate in correspon-
dence with a high prevalence and, vice
versa, of considerable magnitude in
correspondence with a low prevalence,
suggesting that the relative risk was
dependent on the prevalence, which, in
turn, was dependent on disease definition.
The classification of periodontal dis-
ease like any disease is based on
arbitrary or artificial categories con-
structed more for utility than for
Smoking-associated risk 111
naturalness (MacMahon and Trichopou-
los 1996). In the absence of knowledge
of causal criteria, manifestational criter-
ia provide the only basis for categoriza-
tion. Although general acceptance of a
unanimous definition of periodontal
disease does not exist, there is wide
agreement that certain clinical and
radiographic manifestations such as
pocket probing depth, attachment loss,
and bone loss beyond a given limit
(critical level) are diagnostic for the
disease. However, once a decision has
been made on a critical level of the
criterion or set of criteria, the frequency
of the criterion criteria to be required for
the discrimination of disease from non-
disease, i.e., for the classification of an
individual as a case or not, remains to
be decided. In the present study, several
predetermined critical levels based on
various sets of diagnostic criteria were
selected for the definition of disease.
First, critical levels referring to pocket
probing depths of 5 and 6 mm were
alternately used as the diagnostic criter-
ion. Then, for each critical level, several
cutoff points representing increasing
frequencies were selected for the final
definition of disease. In addition, a
certain fraction of the population with
regard to radiographic bone height
distribution, the 30th, 25th, and 20th
percentile, respectively, served the same
purpose. Although arbitrary, these cri-
teria largely concur with those reported
in the literature.
Using a broad definition of disease
such as a cutoff of 1% of pockets with a
probing depth of 5 mm or more, the
overall prevalence for both cohorts
combined was 37% with a prevalence
among smokers that was approximately
twice the prevalence of non-smokers.
This was reflected in an estimated OR
of about 3, suggesting a three-fold
elevated risk for smokers. Using a
narrower definition of disease such as
a cutoff of 15% of pockets with a
probing depth of 5 mm or more, or 10%
of pockets with a probing depth of 6 mm
or more, the overall prevalence de-
creased to approximately 7% and 6%,
respectively. Although the overall pre-
valence related to such a definition
of disease decreased considerably, the
prevalence was 10–15-fold greater
among smokers than non-smokers. This
was reflected in estimated ORs of about
12, suggesting a 12-fold elevated risk
for smokers, i.e., a four times greater
risk than that estimated according to
the first condition. Thus, along with
112 Bergström
gradually more stringent criteria the
overall prevalence decreased whereas
the smoker to non-smoker relative
prevalence gradually increased, which
was reflected in gradually increasing
relative risk estimates. The same pattern
held true regardless of whether the
criterion was selected on the basis of
probing depth, bone height, or combina-
tions of these.
Since the risk estimate catches the
relative prevalence of the exposed ver-
sus nonexposed subpopulations, a good
estimate of the ‘‘true’’ prevalence is
needed. It follows from the above that
since prevalence is related to disease
definition, a ‘‘true’’ prevalence cannot
be unambiguouly determined as long as
the criteria for disease definition vary.
This problem is clearly evident from the
epidemiological literature (e.g., Pilot &
Miyazaki 1991, Brown et al. 1996) and
is further discussed by Burt (1996). A
prevalence of periodontal disease greater
than 10% in the normal adult population,
however, does not seem reasonable and
would imply an unnecessary large pro-
portion of false positives. Applied to the
present observations a ‘‘true’’ prevalence
of 10% would result from a definition of
disease based on pocket frequencies of
10–15% related to 5 mm probing depth,
or 5–10% related to 6 mm probing
depth. Based on such criteria, the mag-
nitude of the smoking-associated rela-
tive risk was in the range 5.6–12.7
(probing depth criteria alone) or 8.3–15.3
(probing depth and bone height criteria
combined). Although, admittedly, the
confidence intervals around these esti-
mates were wide, such a range would
most likely include more precise esti-
mates found from larger sample sizes.
The magnitude of smoking-asso-
ciated risk estimates reported in the
literature varies considerably and so
do the diagnostic criteria. On the basis
of criteria related to pocket probing
depth and attachment loss, respectively,
estimates in the range 1.4–5.3 (Markka-
nen et al. 1985, Hansen et al. 1993,
Stoltenberg et al. 1993, Wakai et al.
1999) and 1.9–9.7 (Locker & Leake
1993, Grossi et al. 1995, Dolan et al.
1997, Paidi et al. 1999, Moore et al.
1999) have been reported. Studies based
on bone height criteria or combinations
of criteria report estimates in the range
1.8–11.8 (Horning et al. 1992, Haber et
al. 1993, Linden & Mullally 1994,
Mullally & Linden 1996, Norderyd &
Hugoson 1998). From these studies it is
apparent that the risk estimate varies
regardless of the diagnostic criterion.
It seems highly likely, therefore, that the
risk estimates reported in the literature
also reflect the cutoff level selected for
disease definition and case identification.
The present observations suggest that
the magnitude of the smoking-asso-
ciated risk is dependent on the magni-
tude of exposure. A positive correlation
between increasing exposure and mag-
nitude of the risk estimate was observed
irrespective of definition of disease and
measure of exposure. Whereas light
exposure was associated with low risk,
the risk associated with heavy exposure
was considerable (9.8–11.4 related to
probing depth alone or 10.7–20.3 re-
lated to probing depth and bone height
combined). The observations are in
accord with the few previous studies
that have estimated the relative risk
associated with exposure gradients
(Grossi et al. 1994, 1995, Haber and
Kent 1992, Haber et al. 1993, Wakai et
al. 1999). The notion that the smoking-
associated risk for periodontal disease
is exposure dependent is important.
Following the concept of causation
advocated by Rothman (1986), light
smoking may not be a component cause
of periodontal disease, whereas heavy
smoking in contrast is highly likely to
be one. Counselling heavy smokers to
reduce smoking (become light smokers)
rather than quit, therefore, may have
substantial import in the context of
prevention.
When estimating the magnitude of
risk associated with tobacco smoking,
several limiting factors have to be
considered such as population charac-
teristics and sample size as well as
methodological limitations such as
smoking misclassification and exposure
inaccuracy. The present study was
based on cross-sectional data from two
samples drawn from the same popula-
tion with an interval of 10 years. The
observations from both time points were
similar, increasing reliability and gen-
eralizability of the observations. Among
the advantages of the study are the
socioeconomical homogeneity of the
population and the high participation
rate (Bergström et al. 1986, 2000a, b).
In addition, the clinical and radio-
graphic data were characterized by high
methodological quality. Among the
limitations are noted the small sample
sizes. As a consequence, risk estimates
were determined with comparably wide
confidence intervals, particularly when
related to exposure substrata. In order to
increase the precision of the estimates
therefore, data from both cohorts were
combined although collected from two
different points in time.
In conclusion, the present observa-
tions suggest that the relative risk
associated with smoking, among other
factors, is dependent on definition of
disease and prevalence. Given other
factors, a narrow definition of disease,
including predominantly aggressive
forms or severe cases, resulted in a
low prevalence and a high risk. For a
broader definition, including also less
distinct forms of disease and borderline
cases, the prevalence increased and the
risk decreased. It was, additionally,
observed that heavy smoking consis-
tently was associated with high risk,
suggesting that for the combination of a
narrow definition of disease and heavy
exposure, the smoking-associated risk
of attracting periodontal disease is
considerable. The results provide at
least in part an explanation to the great
variation reported in the literature con-
cerning the magnitude of risk for perio-
dontal disease attributed to smoking.
Zusammenfassung
Tabakrauchen und Risiko für parodontale
Erkrankungen
Ziele: Die Stärke des Risikos für parodontale
Erkrankungen in Verbindung mit Rauchen
wurde untersucht durch das Erforschen der
Beziehungen zwischen der Definition der Erk-
rankung, der Prävalenz und dem relativen
Risiko in einer Population von zahnbewu!ten
Individuen.
Material und Methoden: Insgesamt 133 Rau-
cher und 242 Nichtraucher im Alter von 20
bis 69 Jahren dienten als Datanquelle. Kriterien,
die auf der klinischen Sondierungstiefe und
den radiographischen Messungen der Knochen-
höhe fu!ten, wurden für den Zweck der
Krankheitsdefinition genutzt. Die verschiede-
nen Frequenzen der Taschen mit den zwei
Sondierungstiefen X5 mm und X6 mm sowie
zusätzlich das 30-, 25- und 20-Perzentil der
Verteilung von der Knochenhöhe waren alter-
native Selektionskriterien.
Ergebnisse: Die Definition der Erkrankung hat
Auswirkung auf die Prävalenz, und das relative
Risiko variiert als eine Funktion der Prävalenz.
Für eine umfassende Definition der Erkrankung
mit 1 % der Taschen X5 mm war die Prävalenz
für Raucher ungefähr zweimal höher verglichen
mit derjenigen von Nichtrauchern. Die Odds-
ratio betrug 3,0. Eine engere Defintion der
Erkrankung mit 15 % der Taschen X5 mm
ergab eine Verringerung der Prävalenz sowohl
für Raucher als auch für Nichraucher und eine
Odds-ratio von 12,1. Die vergrö!erte Exposi-
tion korrelierte mit einer grö!eren Amplitude
des Risikos. Schwere Expositionen waren
mit grö!erem Risiko verbunden als leichtere
Expositionen. Bei einer Kombination einer

engen Definition der Erkrankung und einer
schweren Exposition war das relative Risiko
beträchltich (Odds-ratio 9,8 – 20,3).
Zusammenfassung: Das mit Rauchen verbun-
dene relative Risiko ist abhängig von der
Definition der Erkrankung und der Prävalenz.
Vorausgesetzt andere Faktoren wird eine enge
Defintion in einer niedrigen Prävalenz und
einem hohen Risiko resultieren und umgekehrt
wird bei einer breiten Definition die Prävalenz
hoch und das Risiko moderat sein.
Résumé
Tabagisme et risque de maladie parodontale
Objectif: L’amplitude du risque de souffrir de
maladie parodontale associe au tabagisme a été
recherche en explorant les relations entre la
définition de la maladie, la prévalence et le
risque relatif dans une population d’individus
intéressé par leur dents.
Matériel et Méthodes: 133 fumeurs et 242
non-fumeurs âgés de 20 à 69 ans ont servi de
base de données. Des critères bases sur le
sondage clinique de la poche et des mesures
radiologiques de la hauteur osseuse ont été
utilise pour définir la maladie. Des points
limites variés de fréquence des poches pour
deux niveaux de profondeur de po-
che,4 5 5 mm et4 5 6 mm, et en plus, 30%,
25%, et 20% de distribution de la hauteur
osseuse ont alternativement été utilise comme
critère de sélection.
Résultats: La définition de la maladie a un
impact sur la prévalence, et le risque relatif
varie en fonction de la prévalence. Pour une
définition assez large de la maladie, comme 1%
des poches4 5 5 mm, la prévalence pour les
fumeurs était approximativement le double de
la prévalence chez les non-fumeurs et le odds
ratio était de 3.0. Une définition plus étroite
comme 15% des poches4 5 5 mm résultait en
des prévalences diminuées pour les deux
groupes et un odds ratio de 12.1. Une exposition
plus importante était corrélée avec une plus
grande amplitude de risque. Une lourde exposi-
tion était associe avec un plus grand risque
qu’une exposition légère. Pour une combinaison
d’une définition étroite de la maladie et une
lourde exposition, le risque était considérable
(odds ratio 9.8-20.3).
Conclusion: Le risque relatif associe au
tabagisme dépend de la définition de la maladie
et de la prévalence. D’autres facteurs étant
déterminés, une étroite définition résultera en
une prévalence basse et un risque important, et
vice versa, pour une plus large définition, la
prévalence sera forte et le risque modéré.
References
Bergström, J. (1989) Cigarette smoking as risk
factor in chronic periodontal disease. Com-
munity Dentistry and Oral Epidemiology 17,
245–247.
Bergström, J. & Eliasson, S. (1985) Dental care
habits, oral hygiene, and gingival health in
Swedish professional musicians. Acta Odon-
tologica Scandinavica 43, 191–197.
Bergström, J. & Eliasson, S. (1986) Alveolar
bone height in professional musicians. Acta
Odontologica Scandinavica 44, 141–147.
Bergström, J. & Eliasson, S. (1987a) Noxious
effect of cigarette smoking on periodontal
health. Journal of Periodontal Research 22,
513–517.
Bergström, J. & Eliasson, S. (1987b) Cigarette
smoking and alveolar bone height in subjects
with a high standard of oral hygiene. Journal
of Clinical Periodontology 14, 466–469.
Bergström, J., Eliasson, S. & Dock, J. (2000a)
Exposure to smoking and periodontal health.
Journal of Clinical Periodontology 27, 61–68.
Bergström, J., Eliasson, S. & Dock, J. (2000b)
A 10-year prospective study of tobacco
smoking and periodontal health. Journal of
Periodontology 71, 1338–1347.
Brown, L. J., Brunelle, J. A. & Kingman, A.
(1996) Periodontal status in the United
States, 1988–91: prevalence, extent, and
demographic variation. Journal of Dental
Research 75, 672–683.
Burt, B. A. (1996) Epidemiology of periodontal
diseases. Position Paper. Journal of Period-
ontology 67, 935–945.
Dolan, T. A., Gilbert, G. H., Ringelberg, M.,
Legler, D., Antonson, D., Forster, U. & Heft,
M. W. (1997) Behavioral risk indicators of
attachment loss in adult Floridians. Journal
of Clinical Periodontology 24, 223–32.
Grossi, S. G., Genco, R. J., Machtei, E., Ho, A.,
Koch, G., Dunford, R., Zambon, J. & Haus-
man, E. (1995) Assessment of risk for
periodontal disease. II. Risk indicators for
alveolar bone loss. Journal of Periodontology
66, 23–29.
Grossi, S. G., Zambon, J., Ho, A., Koch, G.,
Dunford, R., Machtei, E., Norderyd, O. &
Genco, R. J. (1994) Assessment of risk for
periodontal disease. I. Risk indicators for
attachment loss. Journal of Periodontology
65, 260–267.
Haber, J. & Kent, R. L. (1992) Cigarette
smoking in a periodontal practice. Journal
of Periodontology 63, 100–106.
Haber, J., Wattles, J., Crowley, M., Mandell, R.,
Joshipura, K. & Kent, R. L. (1993) Evidence
for cigarette smoking as a major risk factor
for periodontitis. Journal of Periodontology
64, 16–23.
Hansen, B. F., Bjertness, E. & Gronnesby, J. K.
(1993) A socio-ecologic model for period-
ontal diseases. Journal of Clinical Period-
ontology 20, 584–590.
Holm, G. (1994) Smoking as an additional risk
for tooth loss. Journal of Periodontology 65,
996–1001.
Ismail, A. I., Morrison, E. C., Burt, B. A.,
Cafesse, R. G. & Kavanagh, M. T. (1990)
Natural history of periodontal disease in
adults: findings from the Tecumseh Period-
ontal Disease Study, 1959–87. Journal of
Dental Research 69, 430–435.
Linden, G. J. & Mullally, B. H. (1994) Cigarette
smoking and periodontal destruction in
young adults. Journal of Periodontology 65,
718–723.
Locker, D. & Leake, J. L. (1993) Risk
indicators and risk markers for periodontal
disease experience in older adults living
independently in Ontario, Canada. Journal
of Dental Research 72, 9–17.
Smoking-associated risk 113
MacMahon, B. & Trichopoulos, D. (1996)
Epidemiology. Principles and methods, 2nd
edition, p. 32. Boston: Little, Brown.
Markkanen, H., Paunio, I., Tuominen, R. &
Rajala, M. R. (1985) Smoking and period-
ontal disease in the Finnish population aged
30 years and over. Journal of Dental
Research 64, 932–935.
Moore, P. A., Weyant, R. J., Mongeluzzo, M.
B., Mayers, D., Rossie, K., Guggenheimer, J.,
Block, H., Huber, H. & Orchard, T. (1999)
Type 1 diabetes mellitus and oral health,
assessment of periodontal disease. Journal of
Periodontology 70, 409–417.
Mullally, B. H. & Linden, G. J. (1996)
Molar furcation involvement associated with
cigarette smoking in periodontal referrals.
Journal of Clinical Periodontology 23,
658–661.
Norderyd, O. & Hugoson, A. (1998) Risk
of severe periodontal disease in a Swedish
adult population. A cross-sectional study.
Journal of Clinical Periodontology 25,
1022–1028.
Norderyd, O., Hugoson, A. & Grusovin, G.
(1999) Risk of severe periodontal disease in a
Swedish adult population. A longitudinal
study. Journal of Clinical Periodontology
26, 608–615.
Paidi, S., Pack, A. R. & Thomson, M. W.
(1999) An example of measurement and
reporting of periodontal loss of attachment
(LOA) in epidemiological studies: smoking
and periodontal tissue destruction. New
Zealand Dental Journal 95, 118–123.
Pilot, T. & Miyazaki, H. (1991) Periodontal
conditions in Europe. Journal of Clinical
Periodontology 18, 353–357.
Preber, H. & Bergström, J. (1986) Cigarette
smoking in patients referred for periodontal
treatment. Scandinavian Journal of Dental
Research 94, 102–108.
Rothman, K. J. (1986) Modern epidemiology,
pp. 7–21. Boston: Little, Brown.
Skaleric, U. & Kovac-Kavcic, M. (2000) Some
risk factors for the progression of periodontal
disease. Journal of the International Acad-
emy of Periodontology 2-1, 19–23.
Stoltenberg, J. L., Osborn, J. B., Pihlstrom, B.,
Herzberg, M., Aeppli, D., Wolff, L. &
Fischer, G. E. (1993) Association between
cigarette smoking, bacterial pathogens, and
periodontal status. Journal of Periodontology
64, 1225–1230.
Wakai, K., Kawamura, T., Umemura, O., Hara,
Y., Machida, J., Anno, T., Ichihara, Y.,
Mizuni, Y., Tamakoshi, A., Lin, Y., Nakaya-
ma, T. & Ohno, Y. (1999) Associations of
medical status and physical fitness with
periodontal disease. Journal of Clinical
Periodontology 26, 664–672.
Address:
Jan Bergström
Odontologiska Institutionen
Box 4064
S-141 04 Huddinge
Sweden