ARTHRITIS & RHEUMATISM

Vol. 42, No. 5, May 1999, pp 882890

1999, American College of Rheumatology

882

FREQUENCY OF FRACTURES IN WOMEN WITH

SYSTEMIC LUPUS ERYTHEMATOSUS
Comparison with United States Population Data
ROSALIND RAMSEY-GOLDMAN, JULIE E. DUNN, CHENG-FANG HUANG, DOROTHY DUNLOP,
JOAN E. RAIRIE, SHIRLEY FITZGERALD, and SUSAN MANZI
Objective. To describe the frequency of selfreported fractures in a large population-based cohort of
women with lupus, to compare the frequency of selfreported fractures between lupus patients and women of
similar age in the general population by use of data
from the 1994 National Health Interview Survey
(NHIS), and to describe the associated risk factors for
fracture in women with lupus. This study is a secondary
analysis of data collected to assess cardiovascular risk
in women with lupus.
Methods. Fractures and associated risk factors
were ascertained by self report in this retrospective
cohort study of 702 living women with lupus who were
followed up for 5,951 person-years. Self-reported fractures were verified in a subset of patients. A Weibull
regression model was used to assess risk factors associated with time from lupus diagnosis to fracture in the
univariate and multivariate analyses. Age-specific standard morbidity ratios (SMRs) were calculated to determine whether fracture occurrence was greater than
expected in women with lupus.
Supported by grants from the NIH, National Institute of
Arthritis and Musculoskeletal and Skin Diseases (AR-30692, AR44811, M01 RR-0056-36, HL-5490002); Commonwealth of Pennsylvania, Department of Health; Lupus Foundation of America, Illinois,
Western Pennsylvania, and Pennsylvania Chapters; Arthritis Foundation Clinical Science grants, the Arthritis Foundation, Illinois Chapter;
the American Heart Association (grant-in-aid); and an unrestricted
educational grant from Merck & Co., Inc.
Rosalind Ramsey-Goldman, MD, DrPH, Julie E. Dunn, PhD,
Cheng-Fang Huang, MS, Dorothy Dunlop, PhD: Northwestern University, Chicago, Illinois; Joan E. Rairie, RN, BSN, Shirley Fitzgerald,
PhD, Susan Manzi, MD, MPH: University of Pittsburgh, Pittsburgh,
Pennsylvania.
Address reprint requests to Rosalind Ramsey-Goldman, MD,
DrPH, Arthritis Ward 3-315, 303 East Chicago Avenue, Chicago, IL
60611.
Submitted for publication August 8, 1998; accepted in revised
form December 9, 1998.

Results. Eighty-six (12.3%) of 702 women reported at least 1 fracture following the diagnosis of
lupus. The sites of the first fracture were the leg (n 5
32), foot (n 5 16), arm (n 5 15), spine (n 5 9), rib (n 5
7), hip (n 5 2), pelvis (n 5 2), hand (n 5 1), shoulder
(n 5 1), and finger (n 5 1). Fracture risk was increased
in the lupus cohort compared with women of similar age
from the United States population, using weighted data
from the 1994 NHIS (SMR 4.7; 95% confidence interval
3.8, 5.8). Variables in the univariate analysis that were
significantly associated (P < 0.05) with time from lupus
diagnosis to fracture were older age at lupus diagnosis,
longer disease duration, longer duration of corticosteroid use, less use of oral contraceptives, and menopause status. In the multivariate analysis, independent
determinants of time from lupus diagnosis to fracture
were older age at lupus diagnosis and longer duration of
corticosteroid use.
Conclusion. Fractures occurred in 12.3% of lupus
patients who were followed up for 5,951 person-years.
There was nearly a 5-fold increase in fracture occurrence in the women with lupus compared with women
from the US population. Older age at lupus diagnosis
and longer use of corticosteroids were associated with
time from lupus diagnosis to fracture. With increased
life expectancy of lupus patients, fracture occurrence is
a major threat to the health of these women. Prevention
strategies must be directed toward minimizing the occurrence of fractures in these patients.
The improved survival of patients with systemic
lupus erythematosus (SLE) has focused attention on the
morbidity associated with the disease and/or its treatment. Women with SLE are at risk of developing low
bone mass, as documented by several cross-sectional

FRACTURES IN WOMEN WITH LUPUS

883

studies (17). However, the frequency of fractures in

women with SLE is unknown.
The pathogenesis of low bone mass and the risk
of fracture in women with lupus are likely to be multifactorial. The relative contributions of traditional or life
style (genetics, race, body weight, exercise, smoking,
alcohol consumption) and SLE-related risk factors (disease activity, disease severity or damage, treatment)
need to be ascertained in order to target intervention
strategies to prevent fractures in these patients.
The objectives of the present study were to
describe the frequency of self-reported fractures in a
large population-based cohort of women with lupus, to
compare the frequency of self-reported fractures between lupus patients and women of similar age in the
general population by use of data from the 1994 National Health Interview Survey (NHIS), and to describe
the associated risk factors for fracture in women with
lupus. This is a secondary analysis of data primarily
collected to assess cardiovascular risk in women with
lupus (8,9).

The 702 women who participated in this study were

more likely to be white (84% versus 62%; P 5 0.001), have
longer lupus disease duration (mean 6 SD 11.0 6 7.1 versus
9.8 6 5.6 years; P 5 0.03), and to be minimally older at entry
(mean 6 SD 45.4 6 13.1 versus 44.9 6 15.5 years; P , 0.01)
and at lupus diagnosis (34.0 6 13.4 versus 33.4 6 16.2 years;
P 5 0.02) than the 207 women who did not participate.
Ascertainment and validation of fractures. Initial patient contact was made at an office visit or via a letter
describing the research study and requesting consent for
participation. Although the primary analysis for the study was
to determine cardiovascular events (8,9), detailed information
was also obtained on fractures occurring before and after lupus
diagnosis. Fracture data were obtained at the time of interview.
In a validation study, a random selection of one-third
of all self-reported fractures were verified by medical chart
review and/or radiology report. Eighty-five percent of these
fractures were confirmed and were not due to severe trauma.
Severe trauma was defined as an injury obtained in a motor
vehicle accident or equivalent event. Fifteen percent could not
be confirmed due to inaccurate patient recall of the date and
place of the fracture or to unavailable medical records and
radiology reports. Self-report is an acceptable method of
ascertaining fracture events (13). Asymptomatic vertebral fractures were underestimated because routine thoracic and lumbar spine radiographs were not obtained for spinal morphometry (14).
Ascertainment of risk factors for fracture. Information
on life style risk factors and lupus-related factors which may be
associated with fractures were collected prospectively for university patients as part of their ongoing participation in the
cohort. This information was updated at the time of the
interview and was obtained for the first time at study interview
for patients being followed up by practicing university rheumatologists in the community. The univariate variables used in
this analysis were smoking, menopause, use of oral contraceptives, use of hormone replacement therapy, pregnancy, lupus
renal disease, age at lupus diagnosis, duration of corticosteroid
use, and average maximum corticosteroid dose.
The life style factors were defined as follows. Smoking
was defined as ever use of tobacco (yes/no) if the women
smoked more than 100 cigarettes in her lifetime. Ever use of
oral contraceptives (yes/no) and estrogen replacement (yes/no)
was obtained. Menopause status (surgical or natural) at the
time of interview was determined by self-report and medical
record review on all study subjects. If menopause status was
unknown, we assumed that the patient was menopausal at age
50, the average age of menopause in the US (15). Standard
hormone evaluations were not routinely performed on patients
in the cohort.
Current age, race, age at lupus diagnosis, and lupus
disease duration from the first physicians diagnosis were
documented. The lupus disease-related factors were renal
disease defined by the ACR criteria for lupus (11) and use of
corticosteroids defined as ever used, duration of use (total
months), and maximum dose. Corticosteroid dose was standardized by conversion to prednisone equivalents (mg/day).
Statistical analysis. The demographic variables and
potential risk factors for fracture were described using descriptive statistics.
For each participant, the person-years of observation

PATIENTS AND METHODS

Lupus population surveyed. The Pittsburgh Lupus
Registry was developed through 2 previous studies and included 916 women who were seen either by staff at the
University of Pittsburgh Medical Center inpatient and outpatient facilities or by practicing university rheumatologists in the
Pittsburgh metropolitan area as of January 1, 1995 (8,9). Data
collection and data management are as previously described
for this cohort (8,10). Briefly, at the time SLE patients enter
the cohort, a standardized data collection form is used to
collect clinical and demographic data. Approximately 450
variables are included on the form, covering demographics,
pertinent aspects of present and past medical history, physical
examination (including reproductive history, the dates of lupus
symptoms onset, physician diagnosis of individual organ system
manifestations, and all of the American College of Rheumatology [ACR] revised classification criteria for lupus [11]),
laboratory evaluations (routine and immunologic), and treatment data, including dose and schedule of corticosteroids and
medications taken for comorbid diseases. All eligible women
18 years of age or older were contacted and asked to participate. This study was approved by our Institutional Review
Board, and all women provided written informed consent.
Seven hundred nine of 916 women from the Pittsburgh
Lupus Registry (77%) agreed to be interviewed for this study.
All patients met the ACR 1982 revised criteria for classification of lupus (11). At the time this study began, the updated
criteria for SLE had not been published (12). Of the 207
nonparticipants, 91 were lost to followup, 58 died, and 58
refused to participate in the study. Seven women were excluded from the analysis because their first post-lupus diagnosis fracture occurred before 1980. The final study population
included 702 women.

884

began at the study interval start date (January 1, 1980) or the

date of the first visit and ended at the time of death, the
occurrence of a fracture, or the end of the study interval
(December 31, 1994). The person-years of observation were
stratified by age groupings (,18, 1824, 2544, 4564, 6569,
and $70 years) to be consistent with the Age Recode 1
groupings in the 1994 NHIS data set (16).
We used standard actuarial methods for determining
fracture incidence adjusted for age (17). For example, if a
woman was age 21 at lupus diagnosis and is currently age 37,
she contributed person-years of observation to 2 different age
strata. The rates were converted to annual incidence rates per
1,000 women with SLE to permit comparison with the NHIS
data. Only the first fracture occurring after lupus diagnosis and
during the observation interval (19801994) was included in
the observed number of fractures.
The 1994 NHIS is a personal interview household
survey using a nationwide multistage sample of 89,100 persons,
designed to represent the civilian, noninstitutionalized population of the US. Health-related events were based on selfreport. Details of NHIS design and implementation are available elsewhere (18). Data on fractures were obtained from
Diagnostic Recode B of the Conditions file. The 47 Recode
variable, which included NHIS Diagnostic Codes 800829, was
used to denote any fracture. The person-years of observation
were stratified into age groups using the Age Recode 1 variable
from the 1994 NHIS (18). The annual final weight variable was
used to generate weighted age and sex-specific annual fracture
rates. Application of appropriate weights and population denominators allows one to generate rate estimates for the US
noninstitutionalized population (National Center for Health
Statistics: personal communication).
Weighted age and sex-specific fracture rates from the
1994 NHIS data were applied to the number of person-years of
observation from the lupus patients in order to generate the
expected number of fractures for this patient group. Agespecific standardized morbidity ratios (SMRs) were calculated
to determine whether fracture occurrence was greater than
expected in women with lupus. The SMRs were calculated by
dividing the observed number of fractures by the expected
number of fractures within each age stratum. Ninety-five
percent confidence intervals (95% CI) were calculated based
on the Poisson distribution for the entire lupus cohort (17).
Risk factors associated with time from lupus diagnosis
to fracture were evaluated using survival analysis methods
based on a Weibull regression model. The Weibull regression
model makes the assumption of proportional hazards. Proportional hazards were tested for the univariate variables, and this
assumption was satisfied for all variables tested (results not
shown). Patients without a fracture contributed censored observations to the analysis. The fit of a Weibull distribution to
the data was assessed through the linearity of the plot of the
complementary log-log of the survival distribution versus the
log of time (19). Univariate Weibull analyses were used to
screen for potential risk factors. The factors that were significantly related to the time from lupus diagnosis to fracture were
entered into a Weibull multiple regression model. KaplanMeier curves were used to illustrate the impact of individual
risk factors (e.g., pre versus post menopause) on the time to
fracture. A Wilcoxon 2-sample test was used to compare the
median age at diagnosis with the disease duration between

RAMSEY-GOLDMAN ET AL

Table 1. Standardized morbidity ratios for fractures in patients with

systemic lupus erythematosus, comparison with United States population estimates from the 1994 National Health Interview Survey*
Age,
years

Person-years
at risk

Observed
no. of
fractures

Expected
no. of
fractures

Standardized
morbidity ratio
(95% CI)

,18
1824
2544
4564
6569
$70
Total

140
587
3,546
1,428
152
98
5,951

2
13
35
32
2
2
86

0.8
1.1
10.9
4.2
0.7
0.4
18.1

2.4 (0.3, 8.7)

12.1 (6.4, 20.7)
3.2 (2.2, 4.4)
7.6 (5.1, 10.7)
2.9 (0.4, 10.5)
4.9 (0.6, 17.7)
4.7 (3.8, 5.8)

* The US population data sample was obtained as part of the National

Health Interview Survey. These data were provided by the National
Center for Health Statistics. The Center specifically disclaims responsibility for any analyses, interpretations, or conclusions. 95% CI 5
95% confidence interval.

groups. Statistical significance was assessed using a nominal a

5 0.05 level of testing.

RESULTS
Fracture information was obtained on 702 living
women from the Pittsburgh Lupus Registry. The study
participants were predominantly white (84%), their median age at lupus diagnosis was 33.2 years, 15% had
renal disease (11), 87% used corticosteroids at some
time during their illness, the median duration of corticosteroid use was 5.0 years, 25% reported smoking, 63%
reported use of oral contraceptives, 22% reported use of
estrogens for hormone replacement therapy, and 81%
had been pregnant at least once. The median time of
observation was 8.2 years; a total of 5,951 person-years
of observation were contributed by the 702 patients.
Thirty-nine percent of the women were classified as
postmenopausal.
There were 86 initial symptomatic fractures
(12.3%) in 702 women after the diagnosis of lupus. The
sites of initial fracture were the leg (n 5 32), foot (n 5
16), arm (n 5 15), spine (n 5 9), rib (n 5 7), hip (n 5
2), pelvis (n 5 2), hand (n 5 1), shoulder (n 5 1), and
finger (n 5 1). The median age at the time of the first
symptomatic fracture was 48.2 years, and the median
number of years from lupus diagnosis to fracture was 5.4
years.
Table 1 shows the age-specific SMRs and 95% CI
for fracture occurrence in the 702 women with lupus
compared with the US population, using data from the
1994 NHIS. The expected number of fractures for the
entire lupus cohort was 18.1; in contrast we observed 86.
The SMR between the observed and expected number

FRACTURES IN WOMEN WITH LUPUS

885

Table 2. Demographic variables and risk factors for time from lupus diagnosis to fracture*

Risk factor
White
Renal disease
Corticosteroid use
Tobacco use
Postmenopause
Oral contraceptives
Estrogen replacement
Pregnancy
Age at lupus diagnosis (years)
Duration of corticosteroid use (years)
Maximum corticosteroid dose (mg)

No. (%)
with fracture
(n 5 86)

No. (%) with

no fracture
(n 5 616)

P
(Weibull model)

73 (85)
9 (15)
78 (91)
20 (24)
46 (53)
43 (50)
25 (29)
75 (87)

515 (84)
78 (15)
527 (86)
152 (25)
225 (37)
396 (64)
127 (21)
497 (81)

0.61
0.56
0.79
0.84
0.02
0.03
0.15
0.25

Fracture
No
event
fracture
(median) (median)

38.9
10
60

32.8
4
60

P
(Weibull model)

,0.0001
0.006
0.18

* Univariate analyses.

of fractures was 4.7, with a 95% CI of 3.8, 5.8. This SMR

indicates that women with lupus were nearly 5 times
more likely to have a fracture compared with similar-age
women from the US population sample. The SMRs for
the various age strata ranged from 2.4 to 12.1. The
largest number of observed fractures and the most years
of observation occurred in the 2544 age stratum (SMR
3.2, person-years of observation after lupus diagnosis
3,546) and the 4564 age stratum (SMR 7.6, personyears of observation after lupus diagnosis 1,428). The
highest SMR occurred in the second-to-youngest age
stratum, 1824, where 1.1 fractures were expected, and
13 fractures were observed.
The univariate results to identify demographic
variables and potential risk factors associated with the
time from lupus diagnosis to first fracture, based on a
Weibull regression survival analysis, are shown in Table
2. Variables in the univariate analysis significantly associated with time from lupus diagnosis to fracture were
older age at lupus diagnosis, longer duration of corticosteroid use, no use of oral contraceptives, and menopause status.
The impact of the age at lupus diagnosis and
duration of corticosteroid use on the time from lupus
diagnosis to fracture is illustrated in Figures 1 and 2.
These plots show Kaplan-Meier curves for women
younger versus those older than the median age at lupus
diagnosis (Figure 1) and for shorter versus longer than
median duration of corticosteroid use (Figure 2). The
time from lupus diagnosis to fracture was not associated
with smoking history, use of estrogen replacement therapy, mean maximum dose of corticosteroids, or lupus
renal disease. Significant risk factors from the univariate

analyses were evaluated in multiple regression Weibull

analyses.
Because age at lupus diagnosis was strongly related to menopause status, these 2 risk factors could not
be simultaneously entered in the same regression model.
Instead, 2 analyses were done, separately entering each
of these 2 risk factors with the remaining factors. The
first multiple regression Weibull analyses showed that
older age at lupus diagnosis (P , 0.001) and duration of
corticosteroid use (P 5 0.004) were related to the time
from lupus diagnosis to fracture, but the use of oral
contraceptives was not related. The second analysis,
which replaced age with menopause status, showed that
duration of corticosteroid use was related to the time

Figure 1. Impact of age on time from lupus diagnosis to fracture. The

median age at the time of lupus diagnosis in women from this cohort
was 33.2 years. Older age was significantly associated with time from
lupus diagnosis to fracture in the univariate analysis (P , 0.001).

886

RAMSEY-GOLDMAN ET AL

Figure 2. Impact of duration of corticosteroid use on time from lupus

diagnosis to fracture. The median duration of corticosteroid use in the
cohort was 5.0 years. Longer disease duration was significantly associated with time from lupus diagnosis to fracture in the univariate
analysis (P 5 0.004).

from lupus diagnosis to fracture; neither menopause

status nor the use of oral contraceptives was related. Of
the 2 analyses, the model including age at lupus diagnosis and duration of corticosteroid use was the stronger
predictor of time to fracture based on the magnitude of
the likelihood functions resulting from the 2 analyses.
To answer the question of whether the age effect
is due to menopause status, we performed a separate
analyses for pre and post menopause status. Since
age was a significant covariate in separate analyses of
pre- and postmenopausal women, this indicates that age
encompasses more information than menopause for
explaining the time from diagnosis to fracture. When the
model included only premenopausal women, age at
lupus diagnosis was significant, P 5 0.001, with time
from lupus diagnosis to first fracture. When the model
included only postmenopausal women, age at lupus
diagnosis was significant, P 5 0.037, with time from
lupus diagnosis to first fracture. These findings are
consistent with our reported results that age at lupus
diagnosis had a probability of P , 0.001 while menopausal status had P 5 0.02. In other words, although age
may be correlated (actually confounded) with menopause, age reflects more information than menopause
status.
DISCUSSION
SLE is a chronic autoimmune disorder that predominantly affects young women. The musculoskeletal

system and bone, in particular, are frequently damaged.

Over 90% of patients have musculoskeletal symptoms at
some time during their disease course. More importantly, 25% of patients have permanent organ damage
attributed to the musculoskeletal system (including osteonecrosis, ruptured tendons, osteoporosis, and vertebral fractures) within the first 10 years of disease (20,21).
Our study is the first to describe the frequency of
self-reported fractures in women with lupus and to
compare the frequency of fractures in these women with
those in women of similar age in the general population.
We have shown that fractures occurred 5 times more
frequently than expected in women with lupus and that
the risk was highest in young women. Almost 50% of the
fractures occurred in women with lupus who were under
the age of 50 or before menopause. In contrast, fractures
are a leading cause of morbidity among postmenopausal
women over age 55 in the general population (22).
An increased frequency of low bone mineral
density (BMD) and reports of fractures in white women
with SLE has been documented in a few small case series
(14,23) and in 2 small cohort studies (refs. 5 and 24 and
Gourley M: personal communication).
Kalla and colleagues (3) using a computer-aided
technique, digitized radiogrammetry, reported increased
metacarpal cortical thinning in 58 women with lupus
compared with 63 normal control subjects. Those investigators extended their observations (4) using a crosssectional study to measure BMD by dual x-ray absorptiometry (DXA) in 108 healthy, menstruating female
volunteers who were not taking medication and 46
menstruating women with lupus without renal disease, of
whom 50% were receiving corticosteroids at the time of
the study. The lupus patients had a mean age of 31 years
and had a mean disease duration of 76 months. Lower
BMD in the femur was noted in the lupus patients
compared with the healthy controls at trabecular (intertrochanter) as well as cortical (femoral neck) sites. The
lupus group as a whole was closer to the lumbar fracture
threshold than was the normal healthy control group.
The prevalence of osteopenia was 25% in the women
with lupus, but was not stated in the controls. No
information on fractures was provided.
Dhillon and coworkers (1) measured lumbar
BMD by DXA in 36 premenopausal women with the
following characteristics: 12 women with lupus who had
been taking at least 10 mg of oral prednisolone for 6
months, 10 women with lupus who had never taken
corticosteroids, 4 women with asthma who had been
taking the equivalent of 10 mg of prednisolone for at
least 6 months, and 10 healthy control women matched

FRACTURES IN WOMEN WITH LUPUS

887

for age. There was no significant difference in mean

lumbar BMD among the 4 patient groups, despite the
higher disease activity, as measured by the University
College Hospital/Middlesex SLE scoring system, and the
higher frequency of menstrual irregularities in the subset
of lupus patients receiving prednisolone.
Formiga and colleagues (2) evaluated BMD by
DXA in a cross-sectional study of 74 premenopausal
white women with lupus compared with 50 premenopausal volunteer control subjects. The mean age (30.8
years) was comparable between the lupus patients and
the control group; the mean disease duration in the
lupus patients was 86.4 months. Twelve percent of the
lupus patients, but none of the controls, had osteoporosis, as defined by the World Health Organization
(WHO) classification (25). No fracture history was
provided. Nevertheless, since low BMD predicts an
increased risk of bone fractures, those authors suggested
that osteoporosis will become a serious problem for
these patients in the future.
Gourley and coworkers (ref. 24 and personal
communication), in a study of 60 patients with lupus,
used DXA to take BMD measurements at the hip and
lumbar spine. This patient group had a mean age of 41
years, and their mean bone density was comparable to
that of historical controls, who were 23 years older than
the patients. The lower BMD in the lupus patients was
not explained by the presence of active versus inactive
disease, duration of lupus (which was used as a surrogate
for the number of years of prednisone treatment), or
presence of nephritis. However, the study may have been
too small to detect clinically significant differences between the subsets of patients with and without these
disease characteristics.
Kipen and colleagues (5) measured BMD at the
lumbar spine and hip in 97 lupus patients (mean age 44.2
years, mean disease duration 8.4 years) from a community setting in Australia. In .40% of the patients, the
BMD at the lumbar spine or hip was rated as osteopenic
by the WHO classification (25), and in 13.4% and 6.3%,
the BMD at the lumbar spine and hip, respectively, was
rated as osteoporotic by the WHO classification (25).
Eleven fractures were reported in this patient group; the
most frequently affected sites were vertebrae and ribs.
There are several theoretical reasons that women
with lupus may have low BMD and increased risk of
subsequent fracture. The consequences of the disease
and/or its treatment that affect traditional osteoporosis
risk factors include 1) irregular menstrual cycles, premature menopause (26), and menopause, 2) avoidance of
hormone replacement therapy (due to concerns about

precipitating a disease flare) (27,28), 3) decreased physical activity (due in part to fatigue) (29), and 4) decreased vitamin D levels (from avoiding sun exposure to
prevent disease flare) (30). The disease-related risk
factors include 1) disease activity due in part to abnormalities in the immune system, such as an imbalance of
inflammatory cytokines including interleukin-6, a cytokine which has been implicated in the pathogenesis of
accelerated bone remodeling (31,32), 2) disease severity
(33), and 3) renal disease (with or without secondary
hyperparathyroidism) (34). Corticosteroid use is the
most frequent treatment-related risk factor (35).
In our study, we addressed some of these issues.
In the univariate analysis, we found that menopause, no
use of oral contraceptives, longer duration of corticosteroid use, and older age at lupus diagnosis were
associated with time from lupus diagnosis to fracture.
Pregnancy, use of estrogens for replacement therapy,
and renal disease were not significant risk factors. Only
older age at lupus diagnosis and duration of corticosteroid use remained significant in the multivariate
model.
In the lupus patients studied by Formiga and
colleagues (2), the mean cumulative dose of prednisone
was 32.5 mg, the average prednisone dosage was 4.5
gm/year, the mean daily dose of prednisone was 13.7 mg
at the time of the BMD measurement, and 72% of the
patients were receiving alternate-day corticosteroid
therapy. There was no correlation between the dosage of
prednisone nor the cumulative dose of prednisone and
BMD at the lumbar spine or femoral neck in the lupus
patients. The lack of correlation between corticosteroid
use and BMD may be explained by the small number of
patients in the study, the early mean age at disease onset
(23 years), the current mean age (31 years), and the use
of alternate-day corticosteroid therapy in 73% of the
patients. In addition, there was no correlation between
disease duration or mean disease activity grade and
BMD in the lupus patients. Similarly, Houssiau and
colleagues (23) found severe trabecular and cortical
bone loss in lupus patients in the total group as well as in
the subset who had been treated with corticosteroids
(23), which supports the hypothesis that osteopenia may
be partly related to the underlying lupus and not just the
corticosteroid treatment.
In contrast, Petris study of patients from the
Hopkins Lupus Cohort (n 5 407) (20) found a strong
association between decreased BMD at the lumbar
spine, as measured by any type of radiography, with both
the cumulative and the highest prednisone dose. Prednisone remained an independent predictor of BMD in a

888

multiple regression model after adjusting for all other

significant predictors of BMD. When patients from the
Hopkins Lupus Cohort were combined with patients
from the University of Maryland, 22 of 364 patients had
sustained a fracture during adulthood. The most common sites of fracture were the hip/femur, vertebra, and
rib. Previous detection of low BMD on any type of
radiograph was one of the predictors of fracture. These
data on fracture occurrence are approximately one-half
the rate reported in the current study, and the reasons
for the discrepancy between the 2 centers are unclear.
In the study by Kipen and coworkers (5), the total
cumulative dose, duration of steroid therapy, peak steroid dosage, and current steroid dosage were all significantly associated with low BMD at the lumbar spine and
femoral neck, even after controlling for age, body mass
index, and disease-related variables (disease activity,
disease severity, and disease duration).
Pons and coworkers (7) compared 43 white premenopausal lupus patients with 43 healthy control subjects. There was no difference in BMD measured at the
lumbar spine and hip between lupus patients and controls. In a post hoc subset analysis, lupus patients were
stratified by prednisone dosage. Patients receiving a
mean prednisone dosage of $7.5 mg/day during their
disease course (mean 4 years) had lower BMD at the
lumbar spine and hip compared with patients who were
receiving ,7.5 mg of prednisone daily or with healthy
controls.
In summary, our study found that the duration of
corticosteroid use was an independent determinant of
time from lupus diagnosis to fracture. This observation is
supported by findings of the Petri, Kipen, and Pons
studies (5,7,20), which suggested that the prednisone
dosage was an important predictor of low BMD. In
contrast, several other studies (1,2,23,24) suggested that
there was no correlation between the cumulative dose of
prednisone and BMD or fractures in predominantly
white premenopausal women with lupus. The reasons
for the conflicting results regarding the importance of
corticosteroids on BMD or fracture in lupus patients
from different centers are unclear, but may relate to
sample size, precision of the measures of corticosteroid
use, duration of SLE, and differences in the mean ages
of the patients.
There are limitations to our study due to its
retrospective design and use of a historical control
group. If menopause status was unknown, we assumed
that the patient was menopausal at age 50, the average
age of menopause in the US (15). Women with lupus are
at risk of premature menopause, and it is possible that

RAMSEY-GOLDMAN ET AL

we misclassified some patients as premenopausal rather

than postmenopausal. Due to the primary interview
strategy used to ascertain cardiovascular events and risk
factors in this cohort, we cannot comment on vitamin D
levels, diet (specifically, daily calcium intake), or family
history of osteoporosis. We could not measure lupus
disease in the laboratory using interleukin-6 (31) or
disease damage using a validated clinical index (33).
The actual fracture event frequencies in our
sample may even be higher than reported because we
were limited in interviewing surviving patients and may
have missed fracture events in those patients who were
deceased. We also did not perform spinal morphometry,
which is the best method for detecting spinal fractures
(14). The prevalence of subclinical bone disease is
likely to be higher since we did not perform bone density
measurements, which is the most sensitive method for
screening for osteoporosis and the best parameter for
predicting fracture risk (36).
The discrepancies between fracture sites in our
patients (most of whom had been treated with corticosteroids) and postmenopausal women may be explained
in part by differences in fracture ascertainment. It is
possible that risk factors other than loss of BMD could
explain the excess of unusual fracture sites in this cohort.
We have information on some, but not all, risk factors
important for bone health. The reasons for excess fractures in this cohort should be assessed in prospective
studies.
We were limited in the availability of comparison
data for fracture occurrence in healthy young women.
We used the 1994 data from the NHIS, which was the
only source that included fracture frequencies in healthy
young adults. Although there could be unforeseen
changes in fracture occurrence in the general population
during the 15-year period covered by the current study
(from 1980 to 1994), the majority of fractures in lupus
patients occurred after 1990. Although the racial distribution (84% white) of women in the Pittsburgh Lupus
Registry and of all participants in the 1994 NHIS is
similar, population denominators used to generate
weighted data stratified by both race and sex were not
available from the NHIS. Therefore, we could not
stratify on both variables simultaneously and cannot
comment specifically on whether fracture occurrence is
increased more than expected in African American
versus white women with lupus. The generalizability of
these data is another potential limitation, although there
is no specific reason to assume that the women with
lupus in the Pittsburgh cohort are unique.
Numerous studies in postmenopausal white

FRACTURES IN WOMEN WITH LUPUS

889

women without lupus have documented the benefit of

various interventions to prevent or decrease fractures if
a woman has low BMD (25). Hormone replacement
therapy (HRT) has been avoided in some patients with
lupus because of concerns about precipitating a disease
flare. However, there is one recent study that supported
the use of HRT in selected patients with lupus (37), and
a clinical trial in progress is evaluating the safety of
estrogen replacement therapy in postmenopausal
women with lupus (38). Prevention strategies directed
toward patients with lupus who are at risk of osteoporosis could include identifying those patients who can
use HRT and those who can use one of the newly
licensed products (bisphosphonates, nasal calcitonin,
and selective estrogen receptor modulators) as alternatives or in addition to HRT. Other prevention strategies
include exercise, calcium and vitamin D supplementation, and strategies to minimize the corticosteroid burden required by these women.
The excess risk of fracture in young women with
lupus is similar to our findings of accelerated cardiovascular disease in the same lupus patient population (8,9).
With the increasing life expectancy of lupus patients, our
studies emphasize the high risk of fracture and cardiovascular disease as major threats to the health of these
women. A prospective study is needed to further define
risk factors for fracture and to design prevention strategies aimed toward minimizing the occurrence of fracture in patients with lupus.

10.

ACKNOWLEDGMENTS

17.

The authors thank Claudia Conte and Claire Dougherty for their assistance with database management and Nichole Katsche for secretarial assistance. We appreciate the
efforts of the Data Management and Methodology Core of the
Multipurpose Arthritis and Musculoskeletal Diseases Center
at Northwestern University Medical School.

18.

REFERENCES

20.

Morand E. Prevalence of reduced bone mineral density in systemic

lupus erythematosus and the role of steroids. J Rheumatol 1997;
24:19229.
Sels F, Dequeker J, Verwilghen J, Mbuyi-Muamba J-M. SLE and
osteoporosis: dependence and/or independence on glucocorticoids. Lupus 1996;5:8992.
Pons F, Peris P, Guanabens N, Font J, Huguet M, Espinosa G, et
al. The effect of systemic lupus erythematosus and long-term
steroid therapy on bone mass in pre-menopausal women. Br J
Rheumatol 1995;34:7427.
Manzi S, Meilahn EN, Rairie JE, Conte CG, Medsger TA Jr,
Jansen-McWilliams L, et al. Age-specific incidence rates of myocardial infarction and angina in women with systemic lupus
erythematosus: comparison with the Framingham study. Am J
Epidemiol 1997;145:40815.
Manzi S, Selzer F, Sutton-Tyrrell K, Fitzgerald SG, Rairie JE,
Tracy RP, et al. Prevalence and risk factors of carotid plaque in
women with systemic lupus erythematosus. Arthritis Rheum 1999;
42:5160.
McCarty DJ, Manzi S, Medsger TA Jr, Ramsey-Goldman R,
LaPorte RE, Kwoh CK. Incidence of systemic lupus erythematosus: race and gender differences. Arthritis Rheum 1995;38:1260
70.
Tan EM, Cohen AS, Fries JF, Masi AT, McShane DJ, Rothfield
NF, et al. The 1982 revised criteria for the classification of systemic
lupus erythematosus. Arthritis Rheum 1982;25:12717.
Hochberg MC. Updating the American College of Rheumatology
revised criteria for the classification of systemic lupus erythematosus [letter]. Arthritis Rheum 1997;40:1725.
Nevitt MC, Cummings SR, Browner WS, Seeley DG, Cauley JA,
Vogt TM, et al. The accuracy of a self-report of fractures in elderly
women: evidence from a prospective study. Am J Epidemiol
1992;135:4909.
National Osteoporosis Foundation Working Group on Vertebral
Fractures. Assessing vertebral fractures. J Bone Miner Res 1995;
10:51823.
Hahn RA. Reliability of reported age at menopause. Am J
Epidemiol 1997;146:7715.
National Center for Health Statistics. National Health Interview
Survey, 1994. CD ROM series 10th ed. Washington (DC): US
Government Printing Office; 1994.
Breslow NE, Day NE. Statistical methods in cancer research: the
design and analysis of cohort studies. Lyon (France): World
Health Organization, International Agency for Research on Cancer; 1987.
National Center for Health Statistics. Design and estimation for
the National Health Interview Survey, 19851994. DHHS publication no. (PHS) 89-1384 ed. Washington (DC): US Government
Printing Office; 1989.
Kalbfleisch JD, Prentice RL. The statistical analysis of failure time
data. New York: John Wiley & Sons; 1980.
Petri M. Musculoskeletal complications of systemic lupus erythematosus in the Hopkins Lupus Cohort: an update. Arthritis
Care Res 1995;8:13745.
Boumpas DT, Fessler BJ, Austin HA, Balow JE, Klippel JH,
Lockshin MD. Systemic lupus erythematosus: emerging concepts.
Part 2. Dermatologic and joint disease, the antiphospholipid
antibody syndrome, pregnancy and hormonal therapy, morbidity
and mortality, and pathogenesis. Ann Intern Med 1995;123:4253.
Melton LJ III, Thamer M, Ray NF, Chan JK, Chesnut CH III,
Einhorn TA, et al. Fractures attributable to osteoporosis: report
from the National Osteoporosis Foundation. J Bone Miner Res
1997;12:1623.
Houssiau FA, Lefebvre C, Depresseux G, Lambert M, Devogelaer
J-P, Nagant de Deuxchaisnes C. Trabecular and cortical bone loss
in systemic lupus erythematosus. Br J Rheumatol 1996;35:2447.
Gourley M, Herion D, Gluck V, Pando J, Yarboro C, Klippel J, et

1. Dhillon VB, Davies MC, Hall ML, Round JM, Ell PJ, Jacobs HS,
et al. Assessment of the effect of oral corticosteroids on bone
mineral density in systemic lupus erythematosus: a preliminary
study with dual energy x-ray absorptiometry. Ann Rheum Dis
1990;49:6246.
2. Formiga F, Moga I, Nolla JM, Pac M, Mitjavila F, Roig-Escofet D.
Loss of bone mineral density in premenopausal women with
systemic lupus erythematosus. Ann Rheum Dis 1995;54:2746.
3. Kalla AA, Kotze TJ, Meyers OL. Metacarpal bone mass in
systemic lupus erythematosus. Clin Rheumatol 1992;11:18.
4. Kalla AA, Fataar AB, Jessop SJ, Bewerunge L. Loss of trabecular
bone mineral density in systemic lupus erythematosus. Arthritis
Rheum 1993;36:172634.
5. Kipen Y, Buchbinder R, Forbes A, Strauss B, Littlejohn G,

6.
7.

8.

9.

11.
12.
13.

14.
15.
16.

19.

21.

22.

23.
24.

890

25.
26.

27.
28.
29.
30.
31.

al. Genetic and metabolic markers of bone mineral density in

systemic lupus erythematosus [abstract]. Arthritis Rheum 1995;38
Suppl 9:S358.
Melton LS. How many women have osteoporosis now? J Bone
Miner Res 1995;10:1757.
Boumpas DT, Austin HA, Vaughn EM, Yarboro CH, Klippel JH,
Balow JE. Risk of sustained amenorrhea in patients with systemic
lupus erythematosus receiving intermittent pulse cyclophosphamide therapy. Ann Intern Med 1993;119:3669.
Jungers P, Dougados M, Pelissier C, Kuttenn F, Tron F, Lesavre
P, et al. Influence of oral contraceptive therapy on the activity of
systemic lupus erythematosus. Arthritis Rheum 1982;25:61823.
Lahita R. The importance of estrogens in systemic lupus erythematosus. Clin Immunol Immunopathol 1992;6:178.
Krupp LB, LaRocca NG, Muir J, Steinberg AD. A study of fatigue
in systemic lupus erythematosus. J Rheumatol 1990;17:14502.
Sontheimer RD, Gilliam JN. Systemic lupus erythematosus and
the skin. In: Lahita RG, editor. Systemic lupus erythematosus. 2nd
ed. New York: Churchill Livingstone; 1992. p. 671.
Linker-Israeli M, Deans RJ, Wallace DJ, Prehn J, Ozeri-Chen T,
Klinenberg JR. Elevated levels of endogenous IL-6 in systemic
lupus erythematosus: a putative role in pathogenesis. J Immunol
1991;147:11723.

RAMSEY-GOLDMAN ET AL

32. Manolagas SC, Jilka RL. Emerging insights into the pathophysiology of osteoporosis. N Engl J Med 1995;332:30511.
33. Gladman DD, Urowitz MB, Goldsmith CH, Fortin P, Ginzler E,
Gordon C, et al. The reliability of the Systemic Lupus International Collaborating Clinics/American College of Rheumatology
Damage Index in patients with systemic lupus erythematosus.
Arthritis Rheum 1997;40:80913.
34. Boumpas DT, Austin HA, Fessler BJ, Balow JE, Klippel JH,
Lockshin MD. Systemic lupus erythematosus: emerging concepts.
Part 1. Renal, neuropsychiatric, cardiovascular, pulmonary, and
hematologic disease. Ann Intern Med 1995;122:94050.
35. Lukert BP, Raisz LG. Glucocorticoid induced osteoporosis: pathogenesis and management. Ann Intern Med 1990;112:35264.
36. Melton LJ, Atkinson EJ, OFallon WM, Wahner HW, Riggs BL.
Long-term fracture prediction by bone mineral assessed at different skeletal sites. J Bone Miner Res 1993;8:122733.
37. Arden NK, Lloyd ME, Spector TD, Hughes GRV. Safety of
hormone replacement therapy (HRT) in systemic lupus erythematosus (SLE). Lancet 1994;3:113.
38. Petri M, Buyon J, Skovron ML, Kim M, for the SELENA Group.
Disease activity and health status (SF-36) in post-menopausal
systemic lupus erythematosus: the SELENA Trial [abstract]. Arthritis Rheum 1997;40 Suppl 9:S208.