We recruited a convenience sample of 320 veterans who
were at least 60 years of age from the dental outpatient
clinic of the Veterans Affairs Hospital, Ann Arbor,
Mich. (n = 206), and from a long-term care facility
(nursing home) associated with the hospital (n = 114).
Patients who use the Veterans Affairs medical and dental
facilities usually have a service-connected injury or
disability, or have met a financial hardship qualification.
There were no age, sex, racial, income or educational
differences among the subjects from the dental outpatient
clinic (that is, the independent living group) and the
long-term care facility (that is, the dependent living
group), which led us to initially combine all the subjects
into a single group for statistical analysis.9
The group from the long-term care facility had significantly
more edentulous subjects than did the outpatient group9;
because of this, a separate variable indicating whether
the participant entered from the independent living
or dependent living situation was included in the multivariate
analysis to account for any confounders that might have
been introduced as a result of entry site.
One of us (B.L.D.) performed a clinical examination
to determine the number of teeth and the number of restorations
and amount of decay on all tooth surfaces in each subject.
The presence and type of dentures and other prosthetic
devices were recorded.
We stratified the subjects into two groups: the first
was composed of subjects with one to 14 teeth and the
second was composed of subjects with 15 to 28 teeth.
This division was done because subjects with one to
14 teeth could have been wearing a full denture, whereas
subjects with 15 to 28 teeth could not wear a full denture.
Third molars, which were rarely present in these older
subjects, were omitted from all analyses.
We measured periodontal pocket depths, attachment levels
and gingival recession for all teeth with an automated
pressure-sensitive periodontal probe (Dental Probe Inc.),
and recorded the results electronically. Oral hygiene
was assessed with the plaque index, or PlI,10
and gingivitis was assessed with the papillary bleeding
score, or PBS.11,12
We calculated the mean PlI and PBS for each subject
by summing the PlIs and PBSs of the individual teeth,
and then dividing the totals by the number of teeth.
These mean PlIs and PBSs were then used in the statistical
analyses. The subjects were asked how often they brushed
and flossed their teeth and how often they visited their
dentist or hygienist. A complaint of xerostomia was
elicited by asking questions about perceived dryness
of the mouth.13
variables. We used stimulated saliva to determine
the number of colony-forming units, or CFU, of selected
bacterial types. The media used, the dispersal procedures
and the culturing conditions have been described elsewhere.14
A curette was used to obtain plaque samples from the
mesial surface of the first molars or, if they were
missing, from the most posterior tooth in each quadrant.
The four plaque samples were individually applied to
the lower reagent strip on the BANA test card. The cards
were incubated at chairside for five minutes at 55 C,15
and the resultant blue color for each plaque sample
was scored and the numbers averaged to give a single
BANA score for each subject.
For this study, CHD had to be a medically established
diagnosis in the patient�s medical record based on the
International Classification of Diseases, 9th Revision,
coding system used by the Veterans Affairs Hospital
for CHD. This diagnosis was supplemented with a review
of the patient�s medical records and documentation of
established myocardial infarction; bypass surgery; clinical
angina; electrocardiogram readings; serum enzyme levels,
if available; angiography; and a positive response to
treatment for heart disease. Systolic and diastolic
blood pressures and blood cholesterol values were obtained
from the patient�s medical records.
We also obtained the subject�s diabetic status from
the medical records; if diabetes was present, the medical
records indicated whether it was controlled by insulin,
diet or medication. Two of the authors (B.L.D. and N.G.)
obtained the patient�s weight and height during the
dental examination, and these measurements were used
to calculate the body mass index, or BMI. We determined
the number of medications used by the subjects through
interviews (performed by B.L.D. and N.G.) and by examining
a computerized record of medications maintained by the
Department of Veteran�s Affairs.
methods. Summary statistics are presented
as means (± standard deviations) and frequencies,
with percentages as appropriate. Initial statistical
tests consisted of X2 analyses and/or
multiple logistic models with Bonferroni-adjusted (P
< .05) pairwise comparisons for categorical data.
After extensive investigation of transformations of
the continuous measures, we concluded that normality
was not achievable, so nonparametric Wilcoxon-Rank sum
or Kruskal-Wallis tests with Bonferroni-adjusted (P
< .05) pairwise comparisons were used. Based on these
initial analyses, variables with P < .25 were entered
into multiple logistic analyses. These models consisted
of predictors with dental relevance and/or statistical
significance. Only significant (P < .05) overall
models were retained, and significant (P < .05) predictors
Two models were developed, one that included all subjects,
but excluded the dental variables (that is, the all-subjects
model) and one that was restricted to the dentate subjects,
but included the dental variables (that is, the dentate
subjects model). In the all-subjects model, we would
be able to observe the effect of variables such as salivary
flow and swallowing, independent of the tooth-related
variables, whereas in the dentate subjects model, any
effect of being edentulous and wearing complete dentures
would not be considered.
These models consisted of predictors with dental relevance
and/or statistical significance, as well as many of
the recognized risk factors for CHD (that is, diabetes,
smoking history, current use of alcohol, age, BMI, blood
pressure and serum cholesterol levels). Based on analyses
of these models, variables with P < .25 were entered
into reduced models until a model was obtained that
combined the highest likelihood ratio with the fewest
degrees of freedom. In this reduced model, the nonsignificant
variables were then added back, one at a time, to the
model containing the significant variables, to determine
their individual effect on the model.
Home | Information | Clinical Studies | Press Releases | Store
123 Primrose Road - Burlingame, CA
Sales: 800-439-2497 Lab: 650-579-1352