By
Joshua J. Filner
Presented to the Department of Public
Health and Preventive Medicine
Oregon Health Sciences University
in Partial Fulfillment of the Requirements
for the Degree of
Master of Public Health
April 2001
School of Medicine
Oregon Health Sciences University
____________________________
Certificate of Approval
____________________________
This is to certify that the M.P.H. thesis of
Joshua J. Filner
has been approved
Thomas
Becker, M.D., Ph.D.
Committee
Chair
Kelly
Krohn, M.D.
Committee
Member
Jodi
Lapidus, Ph.D.
Committee
Member
Table of Contents
List of Tables.................................................................................................................. iv
List of Figures................................................................................................................. v
Acknowledgements..................................................................................................... vii
Funding provided by:.................................................................................................. viii
Abstract............................................................................................................................ ix
Chapter 1 – Introduction............................................................................................... 1
Osteoporosis................................................................................................................. 1
Motivation....................................................................................................................... 4
Chapter 2 – Methods..................................................................................................... 7
Subject Recruitment...................................................................................................... 7
Data Collection Methods.............................................................................................. 8
Data Security and Quality
Control......................................................................... 10
Definitions................................................................................................................ 10
Statistical Methods..................................................................................................... 11
Power Analysis........................................................................................................ 11
Univariate Analysis................................................................................................. 12
Regression Modeling............................................................................................. 12
Chapter 3 – Results..................................................................................................... 14
Chapter 4 – Discussion.............................................................................................. 25
Issues with Ultrasound Bone
Density Technology................................................... 27
Limitations................................................................................................................... 28
Future Studies............................................................................................................. 30
Conclusions................................................................................................................. 31
References..................................................................................................................... 32
Appendix A – Consent Form..................................................................................... 35
Appendix B – The Study Questionnaire................................................................. 37
Appendix C – Methods Notes................................................................................... 39
Appendix D – Ultrasound
Methods.........................................................................
43
Appendix E – Subject T-Score Report......................
List of Tables
4
12
13
18
18
19
19
20
21
22
23
23
24
39
40
List of Figures
Figure 1-1 Normal Bone
Density from Age 30 Years and Older................................. 2
Figure 1-2 W.H.O. Classification
of T-scores................................................................
3
Figure 1-3 Geographical
Distribution of Alaska Natives by Tribal Grouping............. 6
Figure 1-4 Alaska Population:
Age and Sex Distribution by Race.............................. 6
Figure 2-1 Locations
of Study Sites in Alaska...............................................................
7
Figure A-1 Predicted
Values vs. Cook’s Distance from the Final Logistic Regression Model, for
All Natives..........................................................................................................................................
41
Figure A-2 Predicted
Values vs. Pearson Residuals from the Final Logistic Regression Model,
for All Natives..........................................................................................................................................
41
Figure A-3 Predicted
Values vs. Cook’s Distance from a Logistic Regression Model for all Variables
Shown in Table 3-9 and an Age-Smoking Interaction Term, for All Natives..........................
42
Figure A-4 Predicted
Values vs. Pearson Residuals from a Logistic Regression Model for all Variables
in Shown in Table 3-9 and an Age-Smoking Interaction Term, for All Natives.............
42
Figure A-5 Ultrasound
Quality Control Log for August 1999 .....................................
44
Figure A-6 Ultrasound
Quality Control Log for September 1999.............................. 45
Précis
Mitch Greenlick likes
to say that a thesis should be a journey. This thesis certainly has been
a journey, and a circuitous one at that. My first exposure to osteoporosis
was about six years ago; I got a call while still working at the Reed College nuclear
reactor. I spent several months stuffing frozen dead mice into the reactor
to measure calcium levels. That experience ended my promising career as
an animal researcher.
A few years later,
Kelly Krohn sent an e mail to all medical students looking for someone
to go to Alaska and study osteoporosis. I was still interested in osteoporosis and this
project didn’t involve dead mice, so off I went. Like any small student
project, it was unfunded at the start. Fortunately, Dr. Krohn and I were
able to find enough funds to make the project a reality.
The original plan
was that this would be a small summer research project. I was to learn
about field epidemiology, travel, and maybe get published in a scientific
journal. I got to visit some incredibly remote and beautiful places and
meet a lot of wonderful people. At the end of the summer I had some viable
data.
You can always collect
more data. There was also a great deal of interest by both the health
care providers and the patients whom I saw. The project kept growing and
I returned to Alaska. The final result is this document. It is my sincere hope that in the
end it will prove useful in guiding the development of osteoporosis prevention
programs for Alaska Natives.
Acknowledgements
The solo research project in epidemiology
is nearly extinct. This project is no exception and I am extremely grateful
to the many people who helped me. I am unable to list everyone by name. I
apologize to anyone I have omitted.
The nurses, health aides and clerical staff at all of the
clinics deserve special recognition for their help in recruiting, finding
space for me and making me welcome at the clinics. I am most indebted to
the people in the communities I visited, for their hospitality and for
participating in my study.
My Thesis
Committee:
Thomas Becker, M.D., Ph.D., Kelly Krohn, M.D., Jodi Lapidus,
Ph.D.
Logistics,
Advice, and Mentorship:
Jane Kelly, M.D.
Food,
Lodging, Hospitality, Alaska Lore,
and General Support:
Donald, Kara, Kelly, and Lana Johnson, Adrienne Shipley
Proof reading and Editing:
Stephen G. Frantz, P.E., C.H.P.
Christopher Hoffman, M.S.
Encouragement,
Proof reading, and General Tolerance:
Kristina Marie Rheaume
The
Indian Health Service’s Alaska Diabetes
Program Office Staff:
Tammy Brown, R.D., Joan Hastie, Anne-Marie Mayer, F.N.P.,
M.P.H., Anita Vogt, R.N.
Advice:
Kenneth Faulkner, Ph.D., Michael Garland, D.Sci.Rel., Cynthia
Morris, Ph.D., M.P.H., Gary Sexton, Ph.D.
OHSU Bone and Mineral Research Unit:
Eric Orwoll, M.D., Shelia Orwoll
Barrow:
Sharon Conn, R.N., Robert Haight, M.D.
Kotzabue:
Connie Rakoski, Jeanne Reichert, R.N., Janice Shakels,
M.D., Ann Swisher
St.
Paul:
Clinic Staff
Alaska Native Medical Center:
Department of Family
Medicine – Peter Mjos, M.D., Steven Vilter, M.D., Dr. Maria Freeman, M.D.
Department
of Internal Medicine – David Templin, M.D., David Barrett
, M.D.
Department of Orthopedics – David
Beck, M.D., Candice Clawson, M.D., William Pratt, M.D., R.J. Hall, M.D.,
Debbie Freudenthal, R.N
Department of Women’s
Health - Neil Murphy, M.D., Stephanie Ecklund, M.D., Colleen Murphy, M.D.,
Pat Taylor, N.P.
Yukon Kuskaquim Health Corporation:
Roger Johnson, M.D.
Kodiak
Area Native Association:
Dr. Tom Selinger, M.D., Dr. Dave Udder, LT. Felix Santiago,
R.N., B.S.N., U.S.P.H.S.
Sand
Point Medical:
Kathy, Marcy, Margaret, Terry, Ingrid
King
Cove:
Jonathan,
Leslie, Elisabeth Smaha
North Star Health Clinic (Seward):
Annette Siemans, F.N.P., Ellen O’Brian,
F.N.P.
Port Graham Clinic:
Eleanor, Darlene
Tatitlek Clinic:
Laurinda Vlsaoff
Chenga Bay Clinic:
Cheryl Eleshansky
Nanwalek
Clinic Staff:
Effram
Funding
provided by:
The OHSU
Summer Medical Student Research Program (N.I.H.) (1998 and 1999)
The N.L.
Tartar Trust (1999)
The Portland V.A. Research Foundation (1998)
Chugachmuit
Native Corporation (1998, 1999)
North Slope
Native Corporation (1999)
Maniilaq
Native Corporation (1999)
Alaska Area Diabetes Program (1998, 1999)
Aleutian
Pribilof Island Association (1999)
Eastern
Aleutian Tribes
(1998)
The
use of the Ultrasound Device in 1999:
Hologic,
Inc., Waltham,
MA
Abstract
Objective: To
estimate the prevalence of risk factors for osteoporosis in Alaska Natives.
Methods: This study is a cross-sectional
convenience sample of patients from 17 clinics in Alaska. I interviewed
subjects regarding dietary calcium, demographics, and risk factors for
osteoporosis. I measured subjects’ calcaneal bone density using quantitative
ultrasound.
Results: I collected data from 452 women,
316 of whom were Alaska Natives. Risk factors for osteoporosis were highly
prevalent. Many Alaska Natives were current smokers (45%), were former
smokers (32%), and 45% had low bone density (t-score < -1.0). Total
dietary calcium intake was lower than recommended (median = 379 mg/day). Both
current (Odds Ratio (OR)=3.9, 95% Confidence Interval (95% C.I.)=1.8-8.4)
and former smokers (OR=2.8, 95% C.I.=1.3-6.2) were significantly more likely
to have low bone density than subjects who have never smoked. Chronic
users of oral steroids were also significantly more likely (OR=4.7, 95%
C.I.=1.8-12.0) to have low bone density than non-users of oral steroids.
Conclusions: Risk factors for
low bone density and osteoporosis are prevalent in Alaska Native women. The
number of Alaska Native women at risk for osteoporosis will increase during
the next decade due to the aging of this population. A comprehensive prevention
program to reduce the prevalence of modifiable risk factors in this population
is warranted.
Chapter 1 – Introduction
I examined the relationship between risk factors for low bone density
and bone density measurements in Alaska Native women. I implemented a
cross-sectional survey to evaluate a convenience sample of Alaska Native
and non-native women age 20 years and over.
Osteoporosis
Osteoporosis is a
disease in which the normal architecture of bone is disrupted and the matrix
of bone is demineralized. In a healthy adult, two cell types with opposite
functions are active in bone remodeling. Osteoblasts synthesize the organic
components of the bone matrix; over time this matrix mineralizes. Osteoclasts,
by contrast, break down or resorb bone (Junqueira, 1995). When this process
is under normal homeostatic control, the amount of bone resorbed is roughly
equal to the amount of new matrix formed. In osteoporosis, this balance
is disturbed and osteoclasts resorb bone faster than osteoblasts can replace
it. This disruption of remodeling eventually leads to lower bone mass
and a change in bone architecture. These abnormalities reduce bone strength. As
a result, people with osteoporosis are at increased risk for fractures. Most
medical therapies for osteoporosis attempt to restore the balance between
osteoclasts and osteoblasts by inhibiting the osteoclasts. More detailed
discussions of the basic science of bone metabolism can be found in other
sources (Favus, 1996; Riggs, 1995).
Osteoporosis is divided
into two categories, primary and secondary (Rosenberg,
1994). Primary osteoporosis is associated with aging and is more prevalent
with menopause. Secondary osteoporosis is caused by a factor other than
age or menopause. Causes of secondary osteoporosis include Cushing’s syndrome,
type 1 diabetes, alcohol, lithium, anticonvulsants, malabsorption, and
multiple myeloma.
Bone density changes
with age in a characteristic pattern (Figure 1-1). Bone progressively
increases in density from a state of minimal ossification in a newborn
infant to peak density at approximately thirty years of age. After age
thirty, bone mass begins a slow and steady decline until menopause. After
menopause, bone mass declines rapidly (Hansen, 1995). This postmenopausal
bone loss is the reason why osteoporosis prevention efforts are often focused
on women at or just past menopause.
|
Figure 1-1 Normal Bone Density from Age 30 Years and Older
|
|
Osteoporosis is technically a histologic
diagnosis made using a biopsy sample. A biopsy can detect both a decrease
in ossification and interruptions in the normal cross-linked structural network
of bone (Recker, 1996). In practice, biopsy is very rarely used to
make a diagnosis; various bone density measurement techniques are used instead. Bone
density measurements are non-invasive and are almost as accurate in detecting
osteoporosis.
Bone density measurements are interpreted
in three ways, as g/cm
2, as a t-score and as a z-score. A t-score
is the number of standard deviations from the mean density of young, normal,
sex-matched individuals. A z-score is the number of standard deviations from
the mean density for sex and age-matched individuals. The World Health Organization
(W.H.O.) developed the most widely accepted criteria for classifying t-scores
based on measurements taken at the femoral neck or lumbar spine (Kanis, 1994). Under
W.H.O. criteria (Figure 1-2), a t-score greater than ‑1.0 is normal. A
t-score between ‑1.0 and ‑2.5 is classified as osteopenia, or low
bone density. A score of less than ‑2.5 is classified as osteoporosis. These
criteria apply only to measurements taken at the femoral neck or lumbar spine.
The major health consequence
of osteoporosis is increased risk of fracture. A consensus panel study
estimated that 90% of hip and spine fractures in elderly white women are
attributable to osteoporosis (Melton, 1997). Hip and vertebral compression
fractures are the most important morbidities associated with osteoporosis,
but many other types of fractures are associated with this disease. A
national study using Medicare data found 5.3 discharges per 10,000 person-years
with a diagnosis of vertebral fracture for 65 year olds, and 47.8 discharges
per 10,000 person-years with a diagnosis of vertebral fracture for 90 year
olds (Jacobsen, 1992). Hip and vertebral fractures have enormous health
and financial costs. One study estimated the annual U.S. expenditures to treat osteoporotic
fractures to be $13.8 billion in 1995 (Ray, 1997). Hip fractures are associated
with significant mortality. One study found 8% mortality at 3 months post
fracture (Cree, 2000). A nested case control study found 20% mortality
at one year for hip fracture patients at least 70 years old, with a crude
relative risk of 2.4 compared to age matched controls (Wolinsky, 1997). Hip fracture patients were also hospitalized or institutionalized
more frequently after their fracture than age-matched controls. In addition
to traditional outcome measures, all of these fractures can affect quality
of life, activities of daily living, and body image.
Age-associated bone
degeneration does not affect all women equally. The list of known and
putative risk factors for osteoporosis is quite long. Risk factors identified
by the National Osteoporosis Foundation are shown in Table 1-1 (National
Osteoporosis Foundation, 1999). Other putative risk factors for osteoporosis
are under investigation (Melton, 1996). More risk factors are likely to
be identified in the future.
<="502850833">Table
1-1 Selected Risk Factors for Osteoporosis According to the National
Osteoporosis Foundation
Non-modifiable Risk Factors
|
Potentially Modifiable
Risk Factors
|
|
History of Fracture
as an Adult
|
Current Cigarette
Smoking
|
|
History of Fracture
as an Adult in a First Degree Relative
|
Low Body Weight (<127
lb.)
|
|
Caucasian Race
|
Estrogen Deficiency:
-Early Menopause or
Bilateral Oophorectomy
-Premenopausal Amenorrhea
(>1 year)
|
|
Female Sex
|
Lifelong Low Calcium
Intake
|
|
Dementia
|
Alcoholism
|
|
Poor Health/Fragility
|
Recurrent Falls
|
|
|
Inadequate Physical
Activity
|
|
|
Impaired Eyesight
Despite Adequate Correction
|
Motivation
The term “Alaska
Natives” includes three culturally, linguistically and genetically distinct
groups: Eskimos, Indians, and Aleuts. These groups traditionally inhabited
different regions of Alaska (Figure 1-3). Within each of these broad ethnic classifications
exists substantial linguistic and cultural heterogeneity. The diversity
of indigenous Alaskan people limits the value in attempting to make broad
generalizations. Yet studying each subgroup also presents challenges because
of the large number of subgroups and small population in some of the subgroups.
Health care resources
for Alaska Natives have traditionally focused on treating acute medical
conditions. Infectious diseases receive a high level of attention in Alaska
Natives. This is appropriate because of their high incidence. Botulism,
for example, has an incidence of 10.7 per 100,000 person-years in Alaska, substantially higher
than in any other state (Beller, 1998).
Chronic diseases are
beginning to receive more attention because of the realization of their
increasing impact on Alaskan Natives. More Alaska Natives are living to
the older ages associated with many chronic diseases. Osteoporosis is
one such chronic disease that is likely to be increasing among Alaskan
Natives, but few reports have been published on the subject. A search
of MEDLINE using Alaska Natives and osteoporosis or bone density as keywords
of records from 1965 through 1999 yields less than 20 articles. Specialized
references also yield a small number of articles (Fortuine, 1993).
Some evidence suggests
that Alaska Natives may have lower bone density than Caucasians. A study
published by the International Biological Perspective found that Eskimo
people had lower bone density than Caucasians and that the difference between
Eskimos and Caucasians was more pronounced in the elderly (Mazess, 1978). A
small study by the same investigator found that Aleutian Islanders also
have lower bone density than Caucasians (Mazess, 1985).
A large proportion of Alaska Natives in
the baby boom generation is nearing menopause. Figure 1-4 shows the age
structure of the population of the state of Alaska by race. Since osteoporosis is
primarily a disease of post-menopausal women, osteoporosis will become
an important issue in Alaska Native health care with the increase in life
expectancy of Alaska Natives. Information about the distribution of risk
factors in this population will be of use for planning prevention and intervention
programs. My goal for this project was to provide an estimate of the prevalence
of low bone density and the risk factors for low bone density in Alaska
Natives.
Chapter 2 – Methods
This study is a cross
sectional prevalence survey of Alaska Natives. I selected a convenience
sample of patients at 12 sites in Alaska and collected data in Alaska from July 7 to Aug.
31, 1998 and Aug. 4 to Sept.
21, 1999.
Subject Recruitment
|
|
 |
I recruited subjects at a total of 17 Indian
Health Service (I.H.S.) and tribal clinics in 12 municipalities (Figure 2-1). Clinics
were selected as a convenience sample, rather than a random sample. I chose
clinics using the following criteria: 1) approval of local tribal authority,
2) willingness of at least one health care provider to follow-up on patients
at that clinic, 3) the availability of space at the clinic for subject interviews,
and 4) the costs associated with the clinic visit were within the study budget. I
attempted to maximize the geographical diversity of study sites as permitted
by the site selection criteria. I recruited subjects among patients receiving
care from the following clinics: four clinics at the Alaska Native Medical
Center (ANMC) (Family Medicine, Diabetes, Internal Medicine, Orthopedic, and
Women’s Health) and clinics at King Cove, Kodiak, Sand Point, Tatitlek, Nanwalek,
Port Graham, St. Paul,
Seward, Chenega Bay, Barrow, Bethel, and Kotzabue. I also recruited subjects among the Alaska Native and
non-native staff of each clinic. Staff members were recruited either by e
mail (ANMC) or by word of mouth.
I included only
females who were at least 20 years old. Potential subjects were excluded
if they failed to complete the study protocol or if no valid density
measurement could be taken. Clinic staff conducted the initial recruitment
of subjects, either face to face or by telephone. At one village, the
local radio station broadcast a public service announcement to recruit
subjects. All recruiters explained that participation in the study was
entirely voluntary and there was no cost to the subject. No record was
kept of the number of potential subjects who declined to participate
at initial recruitment. When subjects presented for the study interview,
I again explained the voluntary nature of the study and offered to answer
questions about the study. I presented subjects still willing to participate
in the study with a consent form (Appendix A). For those who refused
to participate at the interview stage, I recorded the reason for refusal.
Data Collection Methods
After obtaining
consent, I interviewed all of the subjects using a closed response questionnaire
(Appendix B) at 15 of the clinics. At two clinics (Port Graham and Nanwalek)
the subjects completed the questionnaire themselves.
During the interview,
I entered the subject responses directly into database software (Epi
Info 6.04b, CDC, Atlanta, GA) using a portable computer. Due to time constraints,
subjects in Nanwalek and in Port Graham filled out paper versions of
the questionnaire while waiting to see me. In these cases, I verified
the answers for a sample of questions during the interview. These paper
questionnaires were entered into the database at a later time. I obtained
questionnaire data before the bone density measurement was taken.
I used the questionnaire
data to estimate dietary calcium. During the interview, I asked each
subject how many servings per week in the last year on average they ate
a given food item. Estimating a serving size for each food I determined
the amount of calcium per serving from tables of nutrient values (Appendix
C). I used these values to calculate calcium intake. The calcium
measures are described in the definitions section of this chapter.
I measured calcaneal
bone density using a Sahara Clinical Bone Sonometer (Hologic, Inc., Waltham, MA). I followed
the recommendations of the manufacturer for the use of this device (Appendix
D). The right calcaneus was measured unless the subject had a history
of any injury to the right foot or ankle (except sprains, bunions, or
fracture of the metatarsals or phalanges). If the first measurement
was not satisfactory, I repeated the measurement up to four times. If
none of these repeated measurements passed an automated reliability test,
the mean of the multiple measurements was used as that subject’s bone
density for the statistical analysis. In some cases no valid measurement
was obtained from the subject; these subjects were excluded from the
analysis (n = 6). At the end of the study visit, I provided the subjects
with reports of their density results (Appendix E) and encouraged them
to discuss the results with their health-care provider.
I completed a standard
Indian Health Service Form 803 ambulatory encounter record for every
study visit. This record included the subject’s t-score, estimated bone
density, broadband ultrasound attenuation, speed of sound through bone,
and appropriate notations to aid the interpretation of the results. Records
completed in 1999 were coded as a telephone visit to prevent inadvertent
billing for participation in the study by the hospital. My supervising
physician at each clinic read and signed the encounter records. I placed
the completed encounter record in the subject’s medical chart along with
a photocopy of the consent form.
Data Security
and Quality Control
I verified the questionnaire
responses by comparing selected response variables with the data in individual
electronic medical records. Records for most subjects were available
via direct dial-up access to the computer system at the Alaska Native Medical Center. I
determined from the medication record which subjects were taking the
following drugs: alendronate, etidronate, estrogen,
estrogen/progesterone mixtures, calcium supplements, corticosteroids,
calcitonin, raloxefene, tamoxifen, phenobarbital,
phenytoin and carbamazepine. Birth date and diabetes diagnosis
were also confirmed by electronic medical record.
I encrypted data
files containing identifying information during the course of this study
when not in use. Identifying information such as name and medical record
number were stripped from the electronic data files at the end of the
data analysis to prevent any accidental disclosure of confidential information.
Definitions
The following definitions
were used for the purposes of this study:
Alaska Native:
I considered any person who self-identified as an Alaska Native or American
Indian. This included American Indians living in Alaska but
originally from other states.
Clinic staff: I defined clinic staff as any person who worked at
the clinic or hospital, including cleaning and kitchen staff.
Salpingo-oophorectomy: I considered a subject to have had a salpingo-oophorectomy
if she stated 1) that both of her ovaries were removed, 2) that one ovary
was removed, or 3) hysterectomy had been performed but the subject was
uncertain if her ovaries were still intact.
Calcium Supplement: A calcium supplement was any dietary supplement containing
calcium. This included multivitamins containing calcium.
Diabetic: I considered any person who identified herself as
a diabetic to be a diabetic. Gestational diabetes mellitus and insulin
resistance were not considered diabetes for this study. Subject diabetes
status was confirmed as part of the review of the electronic medical
record.
Menopause: I considered any woman who stated that she had started
or completed menopause to have started menopause. I also considered
any woman who stated she had surgical menopause to have started menopause.
Total
Daily Dietary Calcium: I multiplied the number of servings per
week of each food by the amount of calcium per serving of that food. I
summed the values from all foods and divided by seven.
Total Daily Dairy Calcium:
I multiplied the number of servings per week of each dairy food by the
amount of calcium per serving of that food. I summed the values from
all foods and divided by seven.
Total Daily Calcium from Supplements: I multiplied the number of days per week supplements were used by the
dose of calcium and divided by seven.
Total Daily Calcium Intake:
The sum of total dietary calcium and daily calcium from supplements. This
was the primary calcium measure used in the statistical analysis.
Statistical Methods
Power Analysis
I performed an estimated
power analysis using an estimated prevalence of osteopenia in each group. The
minimum odds ratio to be detected was assumed to be 2.0, alpha was set
to 0.05 and power (1-Beta) to 90% (see Table 2-1). Calculations were
performed using the method described by Fleiss (1981) using Epi Info. The
comparison group consists of non-native people served by the Indian Health
Service and/or the native corporations. These health care systems serve
relatively few non-natives. The non-natives they do serve are mostly
clinic staff and their dependants. Given the small number of non-native
people served by the I.H.S. and the focus of this investigation on Alaska
Natives, I assumed that 1/3 of the subjects would be non-natives. These
results led to the choice of a t-score of –1.0 as the cutoff for a normal
value.
<="502850834"><="475694268">Table 2-1 Estimated Sample Size Given Different Assumptions of Prevalence and Bone
Density Cut Point for Alaska Natives vs. Non-Natives
|
Criteria
|
# Non-Native:
#Native
|
Native
Prevalence
|
Non-Native Prevalence
|
Estimated # Native
|
Estimated #Non-Native
|
Total Subjects
|
|
t-score <=
-
|
1:2
|
0.44
|
0.21
|
129
|
64
|
193
|
|
t-score <=
-2.5
low
prevalence
|
1:2
|
0.10
|
0.005
|
85
|
170
|
255
|
|
t-score <= -2.5 high prevalence
|
1:2
|
0.10
|
0.05
|
368
|
735
|
1,103
|
Alpha = 0.05, Beta = 0.10, Minimum Detectable
Odds Ratio Change = 2.0
Univariate Analysis
I used SPSS 9.0
(SPSS, Inc., Chicago, IL) and Excel 97 (Microsoft, Inc., Redmond, WA) to perform
the statistical analysis. The major variables of interest are shown
in Table 2-2. Bone density was dichotomized into low density (t-score <-1.0)
and high density (t-score ³ -1.0). Dichotomous variables were examined by cross
tabulation with normal vs. abnormal bone density. Likelihood ratio Chi square
statistics and odds ratios (OR’s) were calculated for each of these variables. For
continuous variables I generated descriptive statistics and histograms. In
some cases I assessed the normality of continuous variables with a Kolmogorov-Smirnov
test.
Regression Modeling
I developed a set
of logistic regression models that dichotomized subjects by bone density
to high density and low density using the same criteria used in the univariate
analysis. I used a method, modeled after that of Hosmer and Lemeshow
(1989), to select variables for the multivariate analysis. The
univariate likelihood ratio Chi squared results were used
to screen potential variables in the model. The p-value for these tests
had to be 0.25 or less unless the variable was known from other sources
to have a relationship with osteoporosis. These potential variables
are shown in Table 2-3. I used the gamma statistic to measure the relatedness
of binary variables with one another. Interactions were tested using
both forward and backward stepwise likelihood ratio procedures. I assessed
the fit of the model by plotting the predicted probabilities from the
final models against Cook’s distance and separately plotting predicted
probabilities against the Pearson residuals (Appendix C). Three individual
cases which were outliers on these plots and which changed the estimated
odds ratios of the model were removed from the final analysis. I also
used the Hosmer and Lemeshow goodness of fit test and Nagelkerke R2 as
formal tests of goodness of fit.
Table
2-2 Candidate Predictor Variables and Variable Type, Tested for Use
as Covariates in the Statistical Multivariate Analysis
|
Predictor
|
Variable
Type (units)
|
|
Dietary Calcium Intake from Dairy Products and Supplements
|
Continuous (mg/day)
|
|
Calcium Supplementation
|
Dichotomous
|
|
Clinic Staff vs. Non-Staff
|
Dichotomous
|
|
Native vs. Non-Native
|
Dichotomous
|
|
History of Ankle Fracture
|
Dichotomous
|
|
Physical Exercise of at Least Twenty Minutes per Day 3x per Week
|
Dichotomous
|
|
Oral Corticosteroids for at Least Six Weeks at Any Time in Their Life
|
Dichotomous
|
|
Ever having Smoked Tobacco, Currently Smoking Tobacco,
Not ever Smoking Tobacco
|
Dichotomous
|
|
Cigarette Consumption History
|
Continuous
(packs/day and pack years)
|
|
The use of Hormone Replacement Therapy
|
Dichotomous
|
|
The Use of Oral Contraceptive Pills
|
Dichotomous
|
|
Presence of Diabetes
|
Dichotomous
|
|
Anchorage vs. Other Sites
|
Dichotomous
|
|
Age at Time of Study Measurement
|
Continuous
(years)
|
Chapter 3 – Results
Subjects took between
15-60 minutes to complete the study interview and density measurement. I
interviewed and measured a total of 483 potential subjects, of those 452
(94%) met all inclusion criteria and provided complete data for the purposes
of statistical analysis. The geographical distribution of study subjects
is shown in Table 3-1. Sixteen persons declined to participate in the
study. Three persons declined because they felt they were too young to
have a bone density measurement, three others declined because they did
not want to answer study questions, three did not want to find out if they
had cancer, and two did not wish to potentially lose their insurance coverage. The
remaining five declined to give a reason for not participating in the study.
I reviewed the medical
records of 401 (89%) subjects approximately nine months after the last
study visit. Two persons were deceased at the time of the review. There
were 17 errors in birth dates found in the 401 birth dates reviewed. I
personally entered all of the data, so my error “rate” per keystroke, at
6 keystrokes per birth date, was 0.8%. Only five of these erroneous birth
dates differed by a year or more from the correct birth date. Three subjects
were taking alendronate, four tamoxifen, and one calcitonin. The medical
record agreed closely with the questionnaire data. The study questionnaire
identified diabetics correctly 98% of the time, oral steroid users 93%
of the time, and hormone replacement therapy users 82% of the time.
The primary study
analysis considered risk factors for low bone mass in Alaska Natives (N=316). Risk
factors for osteoporosis were highly prevalent in the study population
(Table 3-2). Smokers made up 35% of all subjects and 45% of Alaska Natives. Nearly
one in five Alaska Natives reported having broken a foot or ankle at some
time in their lives. Ex-smokers comprised 30% of all subjects and 32%
of Alaska Natives. The median tobacco consumption history was eight pack-years. Most
Alaska Native participants (60%) had started menopause and 9% had taken
at least one six week course of oral steroids during their lifetimes. Some
protective factors were also prevalent in the study population. Approximately
one third of the Alaska Natives were taking hormone replacement therapy,
and 49% of Alaska Natives and 61% of non-natives stated that they exercised
at least three times a week for 20 minutes or more per day.
The median total calcium
intake of Alaska Natives was 379 mg/day, and 63% of Alaska Natives reported
that they took some type of dietary supplement containing calcium. The
calcium consumption of all subjects combined was higher (Table 3-3) than
that of Alaska Natives alone. This was due to much higher calcium consumption
by non-native study participants.
Low bone density was
highly prevalent in the study population, as 45% of Alaska Natives and
22% of non-natives had a t-score less than ‑1.0. The mean t-score
of Alaska Natives was ‑0.9 compared to an average t-score of +0.2
for non-natives (Table 3-4).
Many of the risk factors
were significantly associated with low bone density in the univariate analysis. Crude
odds ratios for Alaska Natives and for all subjects are shown in Table
3-5. Fracture history, steroid use, menopause, hormone replacement therapy,
current smoking, former smoking, and age were all positively associated
with low bone density for Alaska Natives. Exercising three times per week
for at least 20 minutes was negatively associated with low bone mass. Variables,
which did not pass the screening criteria for the multivariate model, were
1) clinic staff vs. non-staff, 2) diabetics vs. non-diabetics, and 3) Anchorage vs. all other sites.
Gamma coefficients
for selected variables are shown in Tables 3-6 and 3-7. Diabetes was strongly
associated with the subject having started menopause (gamma = 0.68). The
use of hormone replacement therapy was also strongly related with menopausal
status (gamma=1.0). No other variables were strongly related with one
another.
Several risk factors were significantly associated with low bone density
in the multivariate model (Table 3-8). In the analysis of Alaska Natives,
both current smokers and former smokers were more likely to have low bone
density compared to never smokers. Oral steroid use was also associated
with low bone density (Table 3-8).
The multivariate analysis of all subjects combined showed similar results
(Table 3-9). Smoking history and oral steroid use were less strongly associated
with low bone density than in univariate models. A composite variable
created by combining, history of hip, ankle, or foot fracture, and history
of osteoporosis diagnosis, was significantly associated with low bone density. Alaska
Natives were almost twice as likely to have low bone density than non-natives
(Table 3-9).
I eliminated several variables from the multivariate analysis during statistical
screening. Variables were screened to test a theoretical relationship
between the variable and bone density, or to test for a confounding effect. Diabetes
was evaluated because several of the field visits were made with the diabetes
team. As a result, diabetics were over represented in the study. Diabetic
status did not meet the statistical screening criteria for inclusion in
the model; I concluded it was not an important source of confounding and
eliminated it from the multivariate analysis. Two other variables were
eliminated from the multivariate analysis by this process, being a health
care worker versus all other subjects, and being a subject in Anchorage versus
other sites. Hip fracture, foot fracture, and history of osteoporosis
diagnosis were rare enough events that I chose to combine them into a single
variable in the primary analysis.
The variables that passed the initial screening were included in the initial
multivariate logistic regression models. I included several theoretically
important variables in the final multivariate model even though they were
not statistically significant in that model. Hormone replacement therapy
use and menopause were highly related. When two strongly correlated variables
are included in the same multivariate statistical model they will often
dilute each other’s association with the dependant variable. Normally
two variables that are this strongly correlated would not be placed in
the same multivariate model. In this case, they are both important factors
in explaining the risk for low bone density. Hormone replacement is positively
associated with low bone mass in the univariate analysis. This is true
in the multivariate analysis until age and/or menopause are added to the
model.
The results from the final logistic regression model for Alaska Natives
are shown in Table 3-8. All of the variables in this model were tested
for interactions with age. The age-steroid interaction was significant. The
age-smoking and age-history of fracture interactions approached significance. I
performed fit diagnostics as described in the methods section, on a model
with all of the variables in Table 3-9, and the age-steroid interaction
term. I removed three cases that were highly influential. With the removal
of these cases, the age-steroid term became unambiguously non-significant
(p=0.4), and the age-smoking term became statistically significant. Goodness
of fit diagnostics with the age-smoking term was not as satisfactory. The
Pearson residuals demonstrated greater variance (Figure A-4). I decided
that the best fit and most interpretable model was a model without interaction
terms. It is likely that effect modification exists between two or more
variables in the model but the study sample was too small to draw meaningful
conclusions about the existence of specific relationships. Estimated odds
ratios calculated using the logistic regression model shown in Table 3-8
are shown in Table 3-10.
Table 3-1 Distribution of Study Subjects by Study Site
|
Site
|
Number of Subjects
|
Percent of Total Subjects
|
Population1
|
Percent Population Sampled
|
|
Anchorage
|
184
|
41
|
258,752
|
0.07
|
|
Barrow
|
38
|
8
|
4,397
|
0.86
|
|
Bethel
|
3
|
1
|
5,463
|
0.05
|
|
Chenega Bay
|
10
|
2
|
35
|
29
|
|
King
Cove
|
7
|
2
|
703
|
1
|
|
Kodiak
|
23
|
5
|
6,859
|
0.34
|
|
Kotzabue
|
21
|
5
|
2,964
|
0.71
|
|
Nanwalek
|
20
|
4
|
180
|
11
|
|
Port
Graham
|
31
|
7
|
190
|
16
|
|
Sand
Point
|
18
|
4
|
830
|
2
|
|
Seward
|
64
|
14
|
3,040
|
2
|
|
St.
Paul
|
19
|
4
|
761
|
3
|
|
Tatitlek
|
14
|
3
|
110
|
13
|
|
Total
|
452
|
100
|
284,284
|
0.16
|
1Estimate from Williams, 1999
Table 3-2 Selected Characteristics
of Study Participants
|
Characteristic
|
All Subjects
N=452
|
Native
N=316
|
Non-Native
N=136
|
P-value1
|
|
Current
Smokers
|
35.2%
|
45.3%
|
11.8%
|
<0.001*
|
|
Former
Smokers
|
30.3%
|
31.6%
|
27.2%
|
0.343
|
|
History
of Ankle or Foot fracture
|
18.4%
|
19.0%
|
16.9%
|
0.599
|
|
Oral
Steroid Use
|
8.2%
|
8.9%
|
7.4%
|
0.592
|
|
Started
Menopause
|
58.9%
|
60.4%
|
55.2%
|
0.295
|
|
Hormone
Replacement Therapy
|
33.8%
|
31.3%
|
39.7%
|
0.086
|
|
Exercise
3x/week>=20min
|
52.8%
|
49.4%
|
61.0%
|
0.022*
|
|
Diabetic
|
10.0%
|
10.4%
|
5.9%
|
0.108
|
|
History
of Hip Fracture
|
1.5%
|
1.9%
|
0.7%
|
0.325
|
|
Clinic
Staff
|
35.2%
|
20.3%
|
69.9%
|
<0.001*
|
|
Low
Bone Density
|
30.3%
|
45.3%
|
22.1%
|
<0.001*
|
|
Calcium
Supplement Use
|
58.0%
|
56.0%
|
62.5%
|
0.198
|
|
History
of Hip, or Ankle or Foot Fracture, or Osteoporosis Diagnosis
|
24.6%
|
26.0%
|
21.3%
|
0.290
|
*Significant
for p<0.05
1Native Compared with Non-native Using Likelihood
Ratio Chi Square Test
Table
3-3 Dietary Calcium Measures (mg/day) for All Subjects and for Alaska
Natives
|
Calcium Measure
|
All Subjects
Mean
Median
(25-75 percentile)
|
Alaska Natives
Mean
Median
(25-75 percentile)
|
p-value1
|
|
Dairy Calcium
|
231
140
(39-359)
|
185
107
(24-278)
|
|
|
Total Dietary Calcium
|
360
272
(110-529)
|
306
198
(94-441)
|
<0.001*
|
|
Calcium from Supplements
|
309
0
(0-257)
|
298
0
(0-200)
|
0.028*
|
|
|
669
471
(191-916)
|
604
379
(161-746)
|
<0.001*
|
1Mann-Whitney U test, Alaska Natives vs. Non-Natives
Table
3-4 Bone Density Characteristics and Pack Years of Smoking for Study
Subjects
|
Characteristic
|
All Subjects
|
Alaska Natives
|
Non-native
|
p-value
|
|
Age
(years)
|
49±12
|
50±13
|
51±9
|
0.2091
|
|
t-score
|
-0.6±1.2
|
-0.9±1.2
|
0.2±1.2
|
<0.001*1
|
|
Bone
Mineral Density (g/cm2)
|
0.507±0.136
|
0.483±0.137
|
0.595±0.133
|
<0.001*1
|
|
Median
Pack-Year Smoking History
|
4
|
8
|
0
|
<0.001*2
|
* p <0.05
1Independent Samples t-test
2Mann-Whitney U test
Table
3-5 Univariate Analysis of Subject Characteristics Tabulated by Abnormal
vs. Normal Bone
Density
1
|
Variable
|
All
Subjects
Crude
OR
|
All
Subjects
2
sided
p-value
|
Alaska Native
Crude
OR
|
Alaska Native
2
sided
p-value
|
|
Alaska Native vs.
Non-Native
|
2.9
(1.8-4.6)
|
<0.001*
|
N/A
|
N/A
|
|
History
of Foot or
Ankle
Fracture
|
1.7
(1.0-2.7)
|
0.042*
|
2.3
(1.3-4.1)
|
0.005*
|
|
Exercise
at Least 3x per Week
|
0.5
(0.4-0.8)
|
0.001*
|
0.6
(0.4-1.0)
|
0.016*
|
|
Diabetic
|
1.4
(0.8-2.7)
|
0.270
|
1.3
(0.6-2.7)
|
0.446
|
|
Oral
Steroid Use
|
1.9
(1.0-3.7)
|
0.061*
|
2.8
(1.2-6.4)
|
0.012*
|
|
Started
Menopause
|
3.7
(2.4-5.6)
|
<0.001*
|
4.3
(2.7-7.0)
|
<0.001*
|
|
Hormone
Replacement Therapy
|
1.4
(1.0-2.1)
|
0.085*
|
1.8
(1.1-3.0)
|
0.013*
|
|
Calcium
Supplement Use
|
1.1
(0.8-1.6)
|
0.593
|
1.4 (0.9-2.3)
|
0.174*
|
|
History
of Hip Fracture
|
10.0
(1.2-84.0)
|
0.009*
|
6.2
(0.8-54.0)
|
0.051*
|
|
Clinic
Staff vs. Non-Staff
|
0.6
(0.4-0.8)
|
0.002*
|
0.8
(0.5-1.4)
|
0.404
|
|
Current
Smoker vs.
Never
Smoker
|
2.6
(1.7-4.3)
|
<0.001*
|
2.3 (1.3-4.3)
|
0.005*
|
|
Former
Smoker vs.
Never
Smoker
|
2.1(1.3-3.4)
|
0.004*
|
2.2
(1.2-4.2)
|
0.012*
|
|
History
of Osteoporosis Diagnosis
|
3.4
(1.4-8.3)
|
0.004*
|
2.8
(1.0-7.5)
|
0.036*
|
|
History
of Hip, Ankle, Foot Fracture, or
Osteoporosis
Diagnosis
|
2.3
(1.5-3.5)
|
<0.001*
|
2.7
(1.6-4.6)
|
<0.001*
|
|
Anchorage vs.
All
Other Sites
|
1.2
(0.8-1.8)
|
0.383
|
1.0
(0.6-1.6)
|
0.969
|
|
Age
in 5 Year Intervals2
|
1.4 (1.3-1.5)
|
<0.001*
|
1.4
(1.3-1.6)
|
<0.001*
|
|
Calcium
Intake in mg
|
1.0
(1.0-1.0)
|
0.438
|
1.0
(1.0-1.0)
|
0.061*
|
|
Pack-Years
of Smoking History
|
1.3
(1.2-1.4)
|
<0.001*
|
1.3
(1.1-1.4)
|
<0.001*
|
*p
£0.250 Screening
Level for Multivariate Model
1Abnomal Bone Density is Defined as a t-score of –1.0
or Less by Calcaneal Ultrasound
Table 3-6 Gamma
1 Coefficients
Among Selected Study Variables as an Evaluation of Relatedness of Candidate
Predictor Variables in a Multivariate Analysis for Alaska Native
Subjects Only
|
Variable
|
Started Menopause
Gamma (p-value)
|
Smoking Status3 Gamma
(p-value)
|
Hormone Replacement
Therapy
Gamma (p-value)
|
|
Started
Menopause
|
N/A
|
0.062(0.525)
|
1.0
(<0.001)*2
|
|
Hormone
Replacement Therapy
|
1.0
(<0.001)*2
|
0.054(0.577)
|
N/A
|
|
Oral
Steroid Use
|
0.006(0.975)
|
-0.159(0.348)
|
0.268(0.204)
|
|
Exercise
at Least 3x per Week
|
0.076(0.491)
|
-0.034(0.701)
|
0.250(0.029)
|
|
Smoking
Status3
|
0.062(0.525)
|
N/A
|
0.054(0.577)
|
|
Clinic
Staff vs.
Non-Staff
|
0.013(0.928)
|
-0.343(0.002)
|
0.172(0.250)
|
|
History
of Hip, Ankle, Foot Fracture, or Osteoporosis Diagnosis
|
0.499(<0.001)
|
0.219(0.044)
|
0.296(0.028)
|
|
Diabetes
|
0.688(<0.001)*
|
-0.312(0.018)
|
0.264(0.178)
|
|
Calcium
Supplement Use
|
0.160(0.164)
|
0.078(0.412)
|
0.166(0.171)
|
|
Anchorage vs.
All
Other Sites
|
-0.224(0.065)
|
0.293(0.002)
|
-0.227(0.075)
|
*Highly Related (gamma > 0.6)
1Gamma is Used to Compare the Relatedness of Binary
Variables and is Interpreted Like a Correlation Coefficient
22x2 Table has One Empty Cell Because No Premenopausal
Women were Taking Hormone Replacement Therapy
3Smoking Status Coded as an Ordinal Variable; 0=Never
Smoker, 1=Former Smoker, 2=Current Smoker
Table 3-7 Gamma
1 Coefficients Amoung Selected Study Variables as
an Evaluation of Relatedness of Candidate Predictor Variables in a Multivariate
Analysis for All Study Subjects
|
Variable
|
Started Menopause
Gamma (p-value)
|
Smoking Status3 Gamma
(p-value)
|
Hormone Replacement
Therapy
Gamma (p-value)
|
|
Started
Menopause
|
N/A
|
0.126(0.108)
|
1.00
(<0.001)*2
|
|
Hormone
Replacement Therapy
|
1.00(<0.001)*2
|
-0.010(0.896)
|
N/A
|
|
Oral
Steroid Use
|
0.038(0.825)
|
-0.120(0.401)
|
0.301(0.089)
|
|
Exercise
at Least 3x per Week
|
-0.194
(0.040)
|
-0.186(0.014)
|
-0.097(0.330)
|
|
Smoking
Status3
|
0.126(0.108)
|
N/A
|
-0.10(0.896)
|
|
Clinic
Staff vs. Non-Staff
|
-0.051
(0.608)
|
-0.507(<0.001)
|
-.305(0.002)
|
|
History
of Hip, Ankle, Foot Fracture, or Osteoporosis Diagnosis
|
0.362
(0.001)
|
0.120(0.191)
|
0.239(0.036)
|
|
Alaska Native vs.
Non-Native
|
0.108
(0.297)
|
0.646(<0.001)*
|
-0.181(0.091)
|
|
Diabetes
|
0.639(<0.001)*
|
-0.085(0.440)
|
0.282(0.099)
|
|
Calcium
Supplement Use
|
0.127(0.188)
|
-0.051(0.513)
|
0.149(0.137)
|
|
Anchorage vs.
All
Other Sites
|
-0.221(0.021)
|
0.302(<0.001)
|
-0.214(0.310)
|
*Highly Related (gamma > 0.6)
1Gamma is Used to Compare the Relatedness of Binary
Variables and is Interpreted Like a Correlation Coefficient
22x2 Table has One Empty Cell Because no Premenopausal
Women were Taking Hormone Replacement Therapy
3Smoking Status Coded as an Ordinal; Variable 0=Never
Smoker, 1=Former Smoker, 2=Current Smoker
Table
3-8 Predictors of Low Bone Density from a Multiple Logistic Regression
Model for Alaska Native Subjects Only
|
Variable
|
Adjusted OR2
|
p-value1
|
|
Current
Smoker
|
3.9(1.8-8.4)
|
<0.001*
|
|
Former
Smoker
|
2.8(1.3-6.2)
|
0.010*
|
|
History
of Hip, Ankle, Foot Fracture, or Osteoporosis Diagnosis
|
1.7(1.0-3.1)
|
0.082
|
|
Age
(5 year Increase)3
|
1.5(1.3-1.8)
|
<0.001*
|
|
Hormone
Replacement Therapy
|
0.6(0.3-1.1)
|
0.119
|
|
Started
Menopause
|
1.4(0.6-3.1)
|
0.442
|
|
Oral
Steroid Use
|
4.7(1.8-12)
|
0.001*
|
|
Exercise
at Least 3x per Week
|
0.7(0.4-1.2)
|
0.184
|
*p<0.05
1p-value by Wald Test
2All Variables Listed in the Table are Included in the
Multiple Logistic Regression Model
3Age is Modeled as the Number of 5 Year Intervals from
Age 30
Table
3-9 Predictors of Low Bone Density from a Multiple Logistic Regression
Model for Non Native and Alaska Native Study Subjects
|
Variable
|
Adjusted OR2
|
p-value1
|
|
Current
Smoker
|
2.3(1.3-4.2)
|
0.004*
|
|
Former
Smoker
|
1.6(1.0-2.9)
|
0.103
|
|
History
of Hip, Ankle, Foot Fracture, or Osteoporosis Diagnosis
|
|
0.043*
|
|
Age
(5 year Increase)3
|
1.4(1.2-1.6)
|
<0.001*
|
|
Hormone
Replacement Therapy
|
0.6(0.3-1.0)
|
0.051
|
|
Started
Menopause
|
1.7(0.8-3.3)
|
0.148
|
|
Oral
Steroid Use
|
2.2(1.1-4.8)
|
0.037*
|
|
Exercise
at Least 3x per Week
|
0.7(0.4-1.0)
|
0.056
|
|
Alaska Native vs. Non-native
|
1.8(1.1-3.1)
|
0.024*
|
*p<0.05
1p-value by Wald Test
2All Variables Listed in the Table are Included in the
Multiple Logistic Regression Model
3Age is Modeled as the Number of 5 Year Intervals from
Age 30
Table
3-10 Low Bone Density by Age Group for Alaska Natives
|
Age group (Years)
|
Adjusted OR1
|
|
20-24
|
0.4
|
|
25-29
|
0.7
|
|
30-342
|
1.0
|
|
35-39
|
1.5
|
|
40-44
|
2.3
|
|
45-49
|
3.4
|
|
50-54
|
5.1
|
|
55-59
|
7.8
|
|
60-64
|
12
|
|
|
18
|
1Adjusted for Smoking History, Exercise, Menopausal
Status, Hormone Replacement Therapy Usage, Oral Steroid Use, and Fracture
History
2Reference Group Age 30-34 Years
Chapter 4 – Discussion
Osteoporosis is a
debilitating degenerative disease associated with aging. Environmental,
behavioral, and genetic factors act to either decrease or increase the
risk of osteoporosis. Several factors suggest that Alaska Natives may
have a different risk profile for osteoporosis than the United
States average. Changing behavior
patterns among Alaska Natives combined with an increasing life expectancy
make this an important and interesting population to evaluate for osteoporosis
risk.
Alaska is
exposed to considerably less sunlight than other parts of the United States. Sunlight
is required for the conversion of 7-dehydrocholesterol to previtamin D3,
an essential component for calcium absorption (Marcus, 1996). This conversion
process is less efficient in the elderly, adding to their risk for osteoporosis
and fractures (Heaney, 1996). Inefficient conversion combined with low
exposure to sunlight may have a profound effect on elderly Alaska Natives. The
changing diet of Alaska Native peoples may also increase their risk for
osteoporosis. The diet of indigenous peoples has changed dramatically
in the last 30 years (Whiting, 1998). The effect of these dietary changes
on calcium intake is not well characterized. Informally, I found that
Alaska Natives consume smaller quantities of milk and other dairy products
than the United States average. This
low consumption is due to the high prevalence of lactose intolerance, high
cost, and limited availability of these foods. Lactose intolerance is
a special problem, because the major food items fortified with vitamin
D are dairy foods. This combination of reduced dietary vitamin D intake
and low exposure to sunlight may lower bone density.
I measured an average
calcium intake lower than reported by other researchers. Nobmann and colleagues
(Nobmann, 1992) found a mean dietary intake, excluding dietary supplements,
of 516 mg/day in Alaska Native females and 597 mg/day for Caucasian females
in the National Health and Nutrition Examination Survey II (NHANES II). In
this survey Alaska Natives consumed 306 mg/day on average. When I included
dietary supplements in my sample, the discrepancy between my measurement
and the value reported by Nobmann et. al. disappeared. Cummings and colleagues
(Cummings, 1995) reported a higher mean calcium intake, excluding supplements,
of 713 mg/day in a sample of 9,516 Caucasian women. This informal comparison
supports the hypothesis that Alaska Natives have lower calcium intake than
Caucasian women. All of these estimates are lower than the recommended
dietary intake (1200 mg/day) (National Osteoporosis Foundation, 1999). The
differences between my results and the work of others may be due to the
use of different dietary measures. My dietary survey did not include all
high calcium foods. Notably absent from the measurement was cheese, which
can be a major source of dietary calcium. It is unlikely, but possible,
that including cheese could raise the mean daily intake to the recommended
level. The large difference between measured and recommended intakes suggests
that Alaska Natives are not consuming recommended levels of calcium.
Hormone replacement
therapy and menopause were related to one another and to low bone density. This
confounding relationship is easily explained. Women taking hormone replacement
therapy are older than the study population as a whole. When either menopausal
status or age is added to the statistical model,
this confounding with age is eliminated, and hormone replacement therapy
is then negatively associated with low bone density. An additional
complication is that premenopausal women do not take hormone replacement
therapy, almost by definition. This fact made it impossible for me to
test for statistical interactions between these two important variables
using conventional statistical techniques.
Issues with Ultrasound Bone Density Technology
The diagnosis of osteoporosis
is most often made with a bone density measurement. While the interpretation
of these tests seems simple, several technical issues complicate their
interpretation.
A variety of bone
density measurement techniques are available (Lang, 1991). The most widely
accepted technique is Dual Energy X-ray Absorptiometry (DEXA). DEXA uses
an X-ray source to generate two photon beams of known intensities and energies. The
beams are directed at a specific anatomic site. A detector on the opposite
side of the site measures the amount of radiation that passes through the
tissue. The absorption of radiation roughly follows Beer’s law (Barrow,
1988). Since the intensity of the X-ray beam is known, it is an easy calculation
to determine how much radiation the bone absorbs. Denser bone absorbs
more radiation. The use of two photon energies with different absorption
characteristics allows the device to correct for the amount of soft tissue
surrounding the bone. This correction allows accurate measurements of
relatively deep anatomic sites such as the femoral neck and lumbar vertebrae.
Ultrasound is a relatively
new technology for bone density measurements. Ultrasound methods pass
high frequency sound waves through the anatomic site of interest. Unlike
DEXA, ultrasound cannot correct for the amount of soft tissue around deep
anatomic skeletal sites. Instead of the femoral neck or lumbar vertebrae,
shallow anatomic sites are measured. The most common measurement sites
are the patella, metacarpals, and the calcaneus. Even when measuring at
shallow sites, edema can affect the measurement (Johansen, 1997). Ultrasound
densitometry most commonly takes advantage of two physical properties to
make its measurements, attenuation and speed of sound (Njeh, 1997). Broadband
ultrasound attenuation (BUA) is a measure of the absorption of sound by
bone and is analogous to X-ray absorption used by DEXA. Speed of sound
through the bone is often measured as well. Denser, healthier bone allows
sound waves to propagate at a higher speed than in lower density bone.
DEXA and ultrasound
densitometry are strongly correlated measurements of bone density. Greenspan
and colleagues found correlation coefficients between DEXA at the calcaneus
and several ultrasound units of 0.79-0.86. The relationship was slightly
weaker between DEXA of the femoral neck and ultrasound at the calcaneus
(r = 0.7-0.8) (Greenspan, 1997).
While ultrasound is
a valid technique, it has limitations. Ultrasound is a less precise measurement
technique than DEXA; therefore it has a larger standard deviation. A t-score
of -1.0 obtained by ultrasound is a lower density than a t-score of -1.0
obtained by DEXA. This results in systematic underdiagnoses of osteoporosis
by ultrasound using W.H.O. criteria (Faulkner, 1999).
The differences between
DEXA and ultrasound are the reason I did not use the ultrasound results
as a diagnostic marker for osteoporosis. Using the W.H.O. criteria, of
a t-score less than ‑2.5, to diagnose osteoporosis using calcaneal
ultrasound misclassifies most people with osteoporosis as non-osteoporotic. Theoretically,
I could use a different cut point that adjusts for the differences between
ultrasound and DEXA. No widely accepted method for this adjustment exists. I
opted instead to use a cutoff of -1.0 to define an abnormal result. This
is an accepted W.H.O. criterion for low bone density or osteopenia. Using
this criterion, almost all persons who have osteoporosis will have an abnormal
test result. The disadvantage of this method is that some persons with
low bone mass by DEXA will be misclassified as having normal bone mass. The
cut off of –1.0 strikes a reasonable balance between detecting persons
at high risk of low bone density and avoiding large numbers of false positive
results.
Limitations
This study was a cross
sectional study, so no information about the temporal relationship between
risk factors and bone density was gathered. I selected subjects for the
study by convenience sampling. Subjects are persons using the health care
system, and many of the subjects have reasons to be concerned about osteoporosis. For
example, several subjects were referred to the study because of long term
corticosteroid use.
Communities were selected
based on the level of support from the local health care system and tribal
corporations. These communities may differ from the Alaska Native population
as a whole in terms of availability of health care resources or other factors
that may affect risk factor distribution. I did not visit any sites in Southeast
Alaska, and I did not obtain enough data from northern communities
to test for a relationship between average daylight exposure and bone density.
The calcium measure
I used was, at best, a good estimate of dietary sources of calcium. The
instrument relied heavily on dairy products to estimate calcium intake. This
may not be appropriate because dairy products are only recently widely
available in rural Alaska and are expensive. Many of the study subjects reported
that they were lactose intolerant. These factors imply that dairy calcium
makes up a smaller proportion of total dietary calcium than it does in
the general United States population. The dietary measure also did not account
for the significant seasonal variation in the diet of Alaska Natives (Nobmann,
1992). Foods like salmon, herring eggs, and salmon berries are available
only during a short season during the summer, although some are preserved
and consumed throughout the year. Exercise is also a seasonal activity
in Alaska. The exercise habits of subjects across the seasons
were not assessed in this study.
Misclassification
of ethnicity is a concern in this study. I determined the ethnicity of
subjects by self-report. I did not identify blood quantum. Some non-native
subjects were possibly misclassified as Alaska Natives. This leads to
difficult questions about the definition of an Alaska Native, as intermarriage
between different tribes, and with Caucasians, is quite common. I do not
believe that this limitation seriously compromises my results, because
all subjects were recruited from I.H.S. facilities. Patients at these
facilities must demonstrate some Alaska Native or American Indian heritage
to receive care.
Alcohol abuse is
a risk factor for osteoporosis that I did not attempt to measure. Accurately
assessing alcohol consumption is quite difficult and complicates the patient
interview. I decided that these difficulties outweighed the benefits of
collecting information about alcohol consumption.
In this study, I used
the odds ratio as my measure of association. This measure is statistically
valid, but the odds ratio does not approximate the relative risk in this
case because of the high prevalence of low bone density in this population.
I personally interviewed
the subjects. It is possible that I asked the questions in such a manner
as to bias the results. To minimize this bias I followed the questionnaire
closely and always performed the bone density test after I completed the
interview.
The sample size, while
adequate to test for many main effects, was too small to detect a statistically
significant effect for exercise. A larger sample would make it possible
to more reliably test for effect modification between age and the variables
whose interaction terms approached significance.
Future Studies
Several changes to
this study design should be incorporated into future work. A larger sample,
with better representation of Southeast
Alaska and of native peoples living
above the Arctic Circle, would allow richer statistical models to be constructed. Measurement
of body mass index, serum vitamin D, and better assessment of comorbidities
such as rheumatoid arthritis, cancer, and seizure disorders would make
it possible to adjust statistically for these important factors. A large
enough sample with broad representation would allow for the development
of a reference database of bone densities for this population. A measurement
of the prevalence of lactose intolerant persons would be useful. A further
improvement would be to follow subjects as a cohort to determine the rate
of osteoporotic fractures in this population.
Conclusions
Osteoporosis is
a growing problem for Alaska Natives. My work indicates that several risk
factors for osteoporosis exist in this population. The magnitude
of the associations between risk factors is similar to that seen by other
investigators in cohort studies (Forsen, 1998; Cummings, 1995).
The total calcium
consumption of Alaska Natives in this study is lower than the USDA recommended
level. Given the high prevalence of other risk factors this is a cause
for concern. These results suggest that a population based dietary intervention
is worthwhile. The dietary intervention should not focus only on calcium
intake; instead it should be a comprehensive intervention that emphasizes
a low fat and high fiber diet.
Other modifiable risk
factors should also be addressed in a prevention program. Clearly, a high
priority needs to be placed on smoking cessation programs. The high prevalence
of exercise in the study population is encouraging and suggests that indoor
community centers that facilitate exercise during the winter will be utilized
if built. Programs for the elderly to prevent falls should also be implemented
where feasible. Prevention programs in Alaska need to account for the large seasonal variation in
diet and activity level.
The Alaska Native
population is aging and becoming more at risk for osteoporosis. Interventions
exist which can decrease the risks of developing osteoporosis for individuals
and for whole populations. Effective treatments for osteoporosis are available. Right
now we have a golden opportunity to implement a systematic program to prevent
osteoporosis and fractures. If we take advantage of this opportunity we
can avoid a significant amount of morbidity and mortality in Alaska Natives.
References
Barrow, G. (1988), Physical
Chemistry, 5
th edition, McGraw
Hill, New York
Jacobsen, S.J., et. al.
(1992), Hospitalization with Vertebral Fracture Among the Aged: A National
Population-Based Study, 1986-1989, Epidemiology 3(6):515-518
Johansen, A., Stone, M.D.
(1997), The Effect of Ankle Oedema on Bone Ultrasound Assessment at the
Heel, Osteoporosis International, 7:44-47
Junqueira, L.C., Carneiro,
J., Kelly, R.O. (1995), Basic Histology, 8
th edition, Appleton and
Lang, Stamford, CT
National Osteoporosis Foundation
(1999), Pocket Guide to Prevention and Treatment of Osteoporosis, available
from the National Osteoporosis Foundation, Professional Education Order
Fulfillment, 1150 17
th Street, NW, Suite 500, Washington, D.C.
20036
Njeh, C.F., Boivin, C.M.,
Langton, C.M. (1997), The Role of Ultrasound in the Assessment of Osteoporosis:
A Review, Osteoporosis International, 7:7-22
Nobmann, E. (1993), Nutrient
Values of Alaska Native Foods, available from the Heath Sciences Information
Service, Consortium Library, University of Alaska Anchorage, 3211 Providence
Drive, Anchorage, AK, 99508
Nobmann, E., et. al.
(1992), The Diet of Alaska Native Adults: 1987-1988, American Journal of Clinical
Nutrition, 55:1024-1032
Ray, N.F., Chan, J.K., Thamer,
M., Melton, L.J. (1997), Medical Expenditures for the Treatment of Osteoporotic
Fractures in the United States in
1995: Report from the National Osteoporosis Foundation, Journal of Bone
and Mineral Research, 12(1):24-35
Appendix A – Consent Form
Title: Assessment of Risk Factors for
Osteoporosis Among Alaska Natives
Principle
Investigators: Kelly Krohn, M.D. (503) 494-8963
Joshua Filner, Medical Student (907)
729-1126
Co-investigator: Jane Kelly, M.D. (907)
729-1126
Purpose: You have been invited to participate in this research study because you are a
women over the age of
twenty. The purpose of this research is to learn about the risks
Alaska Native women have for osteoporosis, often called brittle bone disease.
Participation takes about 15 minutes.
Procedures: If you decide to participate in this research study we will ask you question
about what foods you eat and some medical conditions which are known to affect the risk
of getting osteoporosis. Afterwards we will measure how strong your bones are using a
machine approved by the Food and Drug Administration. This machine uses ultrasound,
not x-rays. It takes a
few minutes to set up and about 10 seconds to measure your bone
strength. You will feel the machine buzz while it is working. You will be given a
copy
of the results and a
copy of this form to take home with you. A
doctor or other health care practitioner will discuss any unusual results with you. A
copy or your results will be put
in your medical record. We will also check your medical record to see if you have taken
certain medicines or have certain illnesses known to affect the risk of osteoporosis.
Risks
and Discomforts: There are no known risks associated with the use of this or any
ultrasound. The bone measurement is painless. Some ultrasound gel will get on your
foot. The information we will collect could affect your insurability.
Benefits: You may or may not be personally helped by participating in this study. By
participating in this study you are helping us understand who will get osteoporosis and
help develop programs to prevent it. You also get a
free ultrasound measurement of your
bones which will tell you something about your chances of getting osteoporosis.
Alternatives: You may choose not to participate in this study.
Confidentiality: Neither your name nor your identity will be used for publication or
publicity purposes.
Costs: There is no cost for participation in this study.
Liability: The Oregon Health Sciences University, as a public cooperation, is subject to
the Oregon Tort Claims Act, and self insured for liability claims. If you suffer any injury
from this research project compensation would be offered to you only if you establish
that the injury occurs through the fault of the University, its officers or employees.
However, you have not waived any legal rights by signing this form. If you have further
questions, please call the Oregon Health Sciences University Medical Services Director
at (503) 494-6020 or Dr. David Barrett of the Alaska Native Medical Center at (907)
229-2062. It is not the policy of the U.S. Department of Health and Human Services, or
any federal funding the research project in which you are participating to compensate or
provide medical treatment for human subjects in the event the research results in physical
injury.
Participation: Dr. Kelly Krohn (503) 494-8963 has offered to answer any other
questions you may have about this study. If you have questions regarding your rights as
a
research subject you may contact the Oregon Health Sciences University Institutional
Review Board at (503) 494-7887 or Dr. David Barrett of the Alaska Native Medical
Center Institutional Review Board at (907) 229-2062. You may refuse to participate or
you may withdraw from this study without affecting your relationship with or treatment
at this clinic, Alaska Native Medical Center, or Oregon Health Sciences University.
Your signature below indicates that you have read the foregoing and agree to participate
in this study.
________________ ______________
Study Participant Date
________________ ______________
Investigator Date
Appendix B – The Study Questionnaire
Native Alaskan Bone Mineral
Density Pilot Project
IDENTIFYING
INFORMATION
Id. Number Date
of Interview <mm/dd> Date Entered <today/yy>
Last name <A > First
name & initial <A >
City or town _______________
ZIP #####
Site # ( 1= Anchorage 2=
Sandpoint 3=Kodiak )
Date of Birth <mm/dd/yy> Age
### years Sex <A> (M/F)
Native American <Y> Tribe1_______________
Tribe2 _______________
Have you ever broken
your ankle or foot or had pins placed in them? <Y>
BMD #.### T-Score ##.# BUA ##.# SOS ####.#
Dietary Questionnaire
Will
Multiply servings/week by 4 for monthly
Food Do
you eat Average Servings/week
Milk,fresh (1 cup) <Y> ##.#
Milk, Evaporated(1/4
cup) <Y> ##.#
Milk, powdered <Y> ##.#
(1/3 cup powder w/1 cup
water = 1 cup milk)
Coffee w/milk <Y> ##.#
or cream (1 Tbsp.)
Yogurt, Plain 1 cup <Y> ##.#
Ice Cream (1/2 cup=scoop) <Y> ##.#
Canned Salmon w/bones
(1/4 cup) <Y> ##.#
Sardines (1 can) <Y> ##.#
Fish Head soup (2 cups) <Y> ##.#
Stinkfish? (SERVING SIZE?) <Y> ##.#
Kelp w/herring eggs (1
cup) <Y> ##.#
Small blackfish w/bones
(3 oz.) <Y> ##.#
Dried Salmon, Sockeye
or Red(3 oz.)<Y> ##.#
Whole Needlefish ( 3
oz.) <Y> ##.#
Juices w/Calcium (1 cup) <Y> ##.#
Fishheads (2 heads) <Y> ##.#
Dark Green Leafy Vegetables <Y> ##.#
(1 cup)
Broccoli (2 spears) <Y> ##.#
Cold Cereal (1 1/2 cup) <Y> ##.#
Hot Cereal (1 cup) <Y> ##.#
Egg (1) <Y> ##.#
Do you take Ca supplements
(e.g. tums, Oscal)? <Y>
If so how often/week
# Dose #####
Do you do physical labor
at home or as part of a paid job? <Y>
Do you exercise at least
3 times a week for 20 minutes or more? <Y>
Has a healthcare provider
suggested exercise in the last year or last visit? <Y>
Have you ever broken
your hip? <Y>
Has a health care provider
ever told you you have osteoporosis? <Y>
Do you have diabetes? <Y>
Are you taking or have
you taken steroids(longer than 6 wks.)? <Y>
Do you or have you ever
smoked tobacco? <Y> If so for how long?(years) ##.#
How many packs per day?
#.#
Has any healthcare provider
suggested quitting in the last year? <Y>
For women only:
Have you started menopause
(have your periods stopped)? <Y>
Are you taking hormone
replacement therapy? <Y>
Are you taking or have
you ever taken birth control pills? <Y>
If so for how long?(Years) ##.#
Comments _____________________________________________________________
Appendix C – Methods Notes
Table A-1 Site Specific Recruitment Notes
|
Anchorage
|
In
1998 the investigator worked at several different clinics (Family
Medicine, Internal Medicine, Women’s Health and Orthopedics). Subjects
were referred directly by providers working at that clinic.
In
1999, to fill a shortage of non-native subjects and make more efficient
use of clinic time, appointments were scheduled in advance for many
study subjects in addition to the 1998 methodology. Due to space
constraints and organizational changes subjects were recruited only
from the Orthopedics and the Women’s Health clinics.
|
|
Bethel
|
Very
few subjects met the local institutional requirements for referral
to the study. All subjects had a high degree of suspicion for osteoporosis
(1998
only).
|
|
Kotzabue
Barrow
|
A
traveling rheumatologist referred many of the subjects to the study
(1999
only).
|
|
Sand
Point
Kodiak
|
At
these sites walk-in patients were recruited for the study; no phone
calls were made for recruitment. These visits were made in conjunction
with the diabetes team (1998 only).
|
|
Seward
Tatitlek
Nanwalek
Chenega Bay
Port
Graham
King
Cove
|
At
these sites the majority of subjects were recruited via telephone
calls from the local providers and had appointments scheduled in
advance. Walk in subjects were also inducted into the study.
(Seward
1998 and 1999, Others 1998 only)
|
Table
A-2 Elemental Calcium Content of Food Items per Serving Used in Calculations
of Dietary Calcium Intake1
|
Food
|
Calcium Content per Serving
|
|
Milk
(unknown % fat)
|
297 mg
|
|
Evaporated
Milk
|
164 mg
|
|
Powdered
Milk
|
210 mg
|
|
Coffee
+ Cream
|
22 mg
|
|
Yogurt
|
447 mg
|
|
Ice
Cream (unknown % fat)
|
84 mg
|
|
Cooked
Greens
|
131 mg
|
|
Broccoli
|
36 mg
|
|
Cold
Cereal
|
26 mg
|
|
Hot
Cereal
|
24 mg
|
|
1
egg
|
24 mg
|
|
Ca
fortified juice
|
289 mg
|
|
Sardines
|
351 mg
|
|
Canned
salmon
|
60 mg
|
|
Fish
head Soup
|
168 mg
|
|
Kelp
w/ herring eggs
|
161 mg
|
|
Small
Black Fish
|
236 mg
|
|
Dried
Salmon
|
84 mg
|
|
Whole
Needlefish
|
93 mg
|
1Calcium Values from Nobmann (1993) and using Nutrition
Data System Version 3.2 (Nutrition Coordinating Center of the University of Minnesota, St. Paul, MN)
End of paper.
