Content uploaded by Gunnar Sturfelt
Author content
All content in this area was uploaded by Gunnar Sturfelt on Nov 04, 2014
Content may be subject to copyright.
Risk factors for developing systemic lupus
erythematosus: a case–control study
in southern Sweden
A. A. Bengtsson, L. Rylander
1
, L. Hagm ar
1
,O.Nived
and G. Sturfelt
Departments of Rheumatology and
1
Occupational and Environmental Medicine,
University Hospital of Lund, S-221 85 Lund, Sweden
Abstract
Objective. To explore the risk factors that have been suggested to be associated with the
development of SLE.
Methods. A case–control study was performed and a questionnaire was developed to obtain
the data. Consecutive female incident cases diagnosed between 1981 and 1999 in a defined
geographical area in southern Sweden were included. Controls, matched for calendar year of
birth, were selected randomly from the same area. In total, 85 cases and 205 controls agreed to
participate. The questionnaire included questions about formal education, body weight and
height, medical history, family history of autoimmune diseases, exposure to ultraviolet radiation,
animals, hair-colouring dyes, alfalfa (lucerne) sprouts, smoking and alcohol habits, history of
physical traumata, blood transfusion, silicone breast implants, exogenous oestrogens,
other medication, and significant negative life events.
Results. Using a multivariate model, a history of hypertension wodds ratio (OR) = 3.7, 95%
confidence interval (CI) 1.4–9.8], drug allergy (OR = 3.6, 95% CI 1.4–9.5), a type IuII sun-
reactive skin type (OR = 2.3, 95% CI 1.1–4.8) and a family history of SLE (OR = 6.8, 95%
CI 1.4–32) were all significantly associated with an increased risk of developing SLE, whereas
consumption of alcohol was inversely associated with the risk of SLE (use of alcohol very
seldom, OR = 1.0; 1–150 gumonth, OR = 0.4, 95% CI 0.2–1.0; >150 gumonth, OR = 0.2, 95%
CI 0.1–0.5). A suggested association with increased SLE risk was seen for smoking (OR = 1.8,
95% CI 0.9–3.6) and blood transfusions (OR = 2.3, 95% CI 0.9–5.8). Neither exposure to
exogenous oestrogen nor exposure to hair-colouring dyes was associated with SLE.
Conclusions. Risk factors of both exogenous and endogenous origin were identified in this
population-based series of SLE patients.
K
EY WORDS: SLE, Risk factors, Smoking, Alcohol, Drug allergy, Hypertension, Family history.
Systemic lupus erythematosus (SLE) is an autoimmune
disease characterized by typical involvement of many
different organ systems and by immunological abnorm-
alities, such as hyperactive B cells producing various
autoantibodies. Although genetic and endogenous
hormonal factors are clearly of great importance,
other risk factors, including environmental exposure,
might be equally important in the aetiology of SLE.
Probably many different environmental factors act
together to induce SLE in the genetically predisposed
individual w1x. A number of drugs, especially those
containing aromatic amines, have been suggested as
aetiological factors in SLE w2x. Environmental agents
containing these chemical components, such as tobacco
smoke and hair-colouring products, have therefore
been examined. The studies that have investigated the
effects of hair-colouring products have shown conflict-
ing results w3–6x, while smoking seems to be associated
with an increased risk of developing SLE w7–10x, with
few exceptions w11x. In contrast, recent findings demon-
strate that alcohol consumption is associated with a
decreased risk of SLE w7–9x. Hormonal treatment
has also been studied, and the results indicate that
hormone replacement therapy may be associated with
increased risk of SLE w12, 13x, whereas the use of oral
contraceptives containing oestrogens seems to be
associated with no risk or only a slightly increased
risk of SLE w14–16x. Infectious agents, especially those
of viral origin, have also been discussed as triggers of
Correspondence to: A. Bengtsson.
Submitted 19 April 2001; revised version accepted 29 November
2001.
Rheumatology 2002;41:563–571
ß 2002 British Society for Rheumatology563
SLE for many years w17x. In concordance with these
theories are reports showing that anti-double-stranded
DNA antibodies have been found in dogs owned by
individuals with SLE w18x, and the possibility of agents
transmissible from animals to humans has therefore
been discussed w19x. The role of stressful negative life
events in the onset of autoimmune diseases is con-
troversial. In Graves disease and insulin-dependent
diabetes mellitus, it has been reported that such life
events increase the risk of developing disease w20, 21x,
but this issue has been less investigated in SLE w22x.
The fact that ultraviolet radiation (UVR) can induce
exacerbation of established SLE disease has been
known for a long time w23x, but UVR has been less
examined as a risk factor for the development of
SLE. Additional potential aetiological factors are the
ingestion of alfalfa (lucerne) sprouts w24x and the
occurrence of disorders such as allergy and urticaria
prior to the onset of SLE w16x.
We undertook a population-based case–control study
to investigate these potential risk factors in female SLE
patients from a defined area in southern Sweden and in
controls matched for sex and age.
Materials and methods
Recruitment area
The population of 92 962 adult (>15 yr of age) female
inhabitants in a specific region in southern Sweden
(Lund-Orup Health Care District) was used to retrieve
subjects for the study. A case–control study was
performed within this recruitment area in 1999.
Cases
The cases were a series of female incident patients
retrieved for the period 1981–1999 within a defined
area in southern Sweden. Males were not included
partly because they were too few and partly because
some of the questions were restricted to females. We
retrieved patients by the use of computerized diagnosis
registers at the only hospital in the area (Lund
University Hospital), where both out-patients and
in-patients are registered w25x. Throughout the study
period, the same principle for establishing the diag-
nosis of SLE had been employed: the presence of
multisystemic disorder with concomitant immunological
abnormalities, i.e. confirmed autoimmunity or comple-
ment deficiency, in the absence of any better alternative
diagnosis explaining the patients’ symptoms w26x. After
excluding 22 patients who had died, the total number
of eligible cases was 91. Median age at clinical diag-
nosis was 45 yr (range 16–83) in this group of 91
patients, and median age at diagnosis in the deceased
patients was 63 yr (45–88). The median time from
diagnosis to participation in the present study was 9 yr
(0–18). The median number of American College of
Rheumatology (ACR) classification criteria was 5 (2–10)
in the eligible cases and 6 (3–8) in the deceased patients.
All eligible patients were sent a questionnaire by mail.
Eighty-five (93%) of the patients agreed to participate
in the study. Among the six non-responding patients,
four did not want to participate, one was lost to
follow-up and one had recently had a stroke and
was not able to participate. The median number of
ACR criteria fulfilled was 4 (4–5) in non-responding
patients and 5 (2–10) in responding cases.
Controls
For each case, two female controls were selected
randomly from the general population in the recruit-
ment area using the computerized population register.
Control subjects were matched to cases for calendar
year of birth. The population in the study area is
almost entirely Caucasian, and matching for race was
therefore not done. A questionnaire was sent to all
control subjects. If the questionnaire was not returned
after a reminder telephone call, two new randomly
selected controls were given a questionnaire. In total,
205 controls out of 383 agreed to participate, yielding
a response rate of 53%. The lowest response rate was
observed among the oldest (>75 yr) women. Because
some patients declined to participate, 12 controls were
reallocated according to the matching criteria. In
addition, some questionnaires from controls were
returned late, after two new controls had already been
contacted. This procedure resulted in one matched set
with one control, 55 sets with two controls, 23 sets with
three controls, five sets with four controls and one set
with five controls.
Questionnaire
The questionnaire included questions about formal
education, body weight and height, exposure to UVR
(sun and solarium), animals (ownership of dog and cat,
and close contact with horse, sheep, goat and other),
hair-colouring dyes (frequency and colour), diet
(consumption of alfalfa sprouts), smoking (number of
cigarettes per day, years of starting and stopping),
alcohol consumption (quantity), history of serious
accidents and blood transfusion, silicone breast
implants, hormonal factors (postmenopausal hormone
replacement therapy, use of oral contraceptives contain-
ing oestrogen, miscarriages, age at menarche and
menopause), medication with specific drugs (sulphasa-
lazine, penicillamine and gold salts). Sun-reactive
skin types were classified according to the Fitzpatrick
skin typing system: individuals reporting ‘always burn,
never tan’ after exposure to the sun were classified as
skin type I, those reporting ‘never burn, always tan’
as skin type IV, and intermediates as skin type II or III
w27x. A history of diseases other than SLE, including
drug allergy, and a family history of autoimmune
diseases were also included in the questionnaire. In
addition, questions about significant negative life events,
such as divorce, loss of a close family member,
unemployment etc., were asked.
Importantly, for all cases only exposures and
histories preceding the time of clinical diagnosis were
used in the statistical analyses. The exposures for the
564 A. A. Bengtsson et al.
controls were also estimated for the period preceding
the year of diagnosis of their corresponding case. In
order to achieve this, the questions in the questionnaire
were generally constructed in such a manner that the
participants were first asked about the present expos-
ure, and were then asked if this had been different
previously. If no change in exposure was reported,
the present exposure was used in the analyses as this
would correspond to exposure before disease onset.
If a different exposure was reported during any
previous period, the participant was asked to specify
this period (more than one period was possible) and
the quantity of exposure. As described above, the year
of diagnosis (i.e. disease onset) was taken into account,
so that only the data from before disease onset were
used. If more than one level of exposure was given for
this period, the maximum level of exposure reported
was used. An example showing the structure of one of
the questions in the questionnaire is given in Fig. 1.
The Ethics Committee of Lund University approved
the study.
Data analysis
The effect of the exposure variables on SLE were
measured by means of the odds ratio (OR) and its 95%
confidence interval (CI). Estimations were performed
by conditional logistic regression w28x. In the multi-
variate analyses we also tested whether effect modifica-
tion was present by including relevant interaction terms
in the models.
Results
UVR
Data on UVR-related risk factors are summarized in
Table 1. Women with the skin type ‘always burn,
neverusometimes tan’ (sun-reactive skin type IuII) had
an increased risk of developing SLE compared with
women with other skin types. A similar finding was
observed for subjects reporting marked burning after
sun exposure during youth (i.e. when aged <20 yr).
Smoking, alcohol and body mass index
A negative dose–response relationship between con-
sumption of alcohol and the risk of SLE was observed
(Table 2). The OR was 0.4 for those with an alcohol
consumption >150 gumonth, and in this group the
median consumption was 307 gumonth. Smokers tended
to have a greater risk of SLE compared with non-
smokers (OR = 1.5, 95% CI 0.9–2.6). Women with a
body mass index (BMI) greater than 21 kgum
2
tended
to have a greater risk of SLE compared with
women with lower BMI (OR = 1.5, 95% CI 0.8–2.8).
However, neither smoking habits nor BMI showed any
dose–response relationship (Table 2).
Self-reported autoimmune diseases in relatives
The proportions of autoimmune diseases in first-degree
relatives and the corresponding ORs are summarized
in Table 3. A family history of any autoimmune
disease was associated with increased risk of SLE.
Especially a family history of SLE or multiple sclerosis
was associated with a substantially increased risk
of SLE.
Self-reported diseases and medication
The proportions of self-reported diseases and the
corresponding ORs are summarized in Table 4.
Hypertension was associated with increased risk of the
development of SLE. Among the infectious diseases,
only pneumonia was reported more often by cases than
FIG. 1. An example showing the structure of one of the
questions included in the questionnaire. The questions were
generally constructed in such a manner that the participants
were first asked about their present exposure, and were then
asked if their exposure had been different previously. If no
change in exposure was reported, the data were used in the
analyses as the information would indicate exposure before
disease onset. If a different exposure was reported during any
previous period, the participant was asked to specify this
period (more than one period was possible) and the quantity of
exposure. The year of diagnosis (i.e. disease onset) was taken
into account, so that that only the data relating to the period
before disease onset were used. If more than one level of
exposure was given in this period, the maximum level of
exposure reported was used.
565Risk factors for developing SLE
by controls. A history of blood transfusion was asso-
ciated with a barely significantly increased risk of
SLE. The reason for transfusion was almost exclusively
surgery. Cases reported gastric ulcer more often than
controls, whereas cancer was reported more often
by controls, but these differences were not statistically
significant. A history of drug allergy was significantly
associated with increased risk of SLE (OR = 2.6, 95%
CI 1.2–2.9; data not shown in Table 4). Very few cases,
and none of the controls, reported medication with
drugs, such as sulphasalazine, penicillamine and gold
salts (data not shown).
Hormonal factors
Age at menarche, age at menopause and numbers
of pregnancies and miscarriages did not differ
between cases and controls (data not shown). A rela-
tively high proportion of cases (48%) had used
TABLE 1. Distribution of cases and controls according to UVR-related risk factors, with corresponding OR and 95% CI
Cases Controls
Variable n (%) n (%) OR 95% CI
Skin type
III–IV: sometimesunever burn, always tan 46 (54) 156 (76) 1.0 (Reference group)
I–II: always burn, neverusometimes tan 39 (46) 49 (24) 2.9 1.6–5.1
Seriously burnt by the sun as a youth (<20 yr of age)
Never 19 (23) 72 (36) 1.0 (Reference group)
More than once 64 (77) 128 (64) 2.2 1.2–4.1
TABLE 2. Distribution of cases and controls according to BMI and exposure to alcohol and smoking, with corresponding OR and 95% CI
Cases Controls
Variable n (%) n (%) OR 95% CI
Alcohol consumption (gumonth)
0 33 (39) 54 (26) 1.0 Ref
>0–150 32 (38) 73 (36) 0.7 0.3–1.3
>150 20 (24) 78 (38) 0.4 0.2–0.8
Smoking (pack yr)
0 34 (40) 100 (49) 1.0 Ref
>0–10 24 (28) 51 (25) 1.5 0.8–2.7
>10 27 (32) 54 (26) 1.5 0.8–2.9
BMI
(20.6 22 (28) 67 (34) 1.0 Ref
>20.6–22.9 30 (38) 66 (34) 1.4 0.7–2.9
>22.9 28 (35) 63 (32) 1.6 0.8–3.2
TABLE 3. Distribution of cases and controls according to the existence of self-reported first-degree relatives with autoimmune diseases, with
corresponding OR and 95% CI
Cases Controls
Disease in first-degree relative n (%) n (%) OR
a
95% CI
Any autoimmune disease 45 (53) 80 (39) 1.7 1.0–2.9
Diabetes type I 2 (2.4) 3 (1.5) 1.9 0.3–11
Multiple sclerosis 5 (6.0) 2 (1.0) 6.0 1.2–31
Goitre 18 (21) 33 (16) 1.4 0.7–2.7
Inflammatory bowel disease 4 (4.8) 7 (3.5) 2.3 0.6–9.3
Gluten enteropathy 1 (1.2) 3 (1.5) 1.2 0.1–11
Psoriasis 10 (12) 25 (12) 1.0 0.4–2.1
Psoriasis arthropathy 3 (3.5) 4 (2.0) 2.1 0.5–9.7
Rheumatoid arthritis 15 (18) 20 (10) 1.9 0.9–3.8
SLE 7 (8.3) 4 (2.0) 4.8 1.2–19
Sjo¨gren syndrome 4 (4.8) 3 (1.5) 3.8 0.7–22
Systemic sclerosis 1 (1.2) 1 (0.5) 2.8 0.2–47
Ankylosing spondylitis 1 (1.2) 3 (1.5) 0.8 0.1–7.8
a
Women without the present disease in a first-degree relative were used as the reference category.
566 A. A. Bengtsson et al.
oestrogen-containing oral contraceptives, but the pro-
portion was even higher in controls (53%). The use of
postmenopausal hormone replacement therapy was less
frequent (16% of cases and 12% of controls). Only five
women (two cases and three controls) reported hormo-
nal treatment in association with in vitro fertilization,
and only four (three cases and one control) had silicone
breast implants.
TABLE 4. Distribution of cases and controls according to presence of self-reported diseases, with corresponding OR and 95% CI
Cases Controls
Disease n (%) n (%) OR
a
95% CI
Atherosclerosis-related diseases
Myocardial infarction 2 (2.4) 1 (0.5) 4.0 0.4–44
Stroke 3 (3.5) 3 (1.5) 2.4 0.5–12
Intermittent claudication 2 (2.4) 2 (1.0) 3.2 0.3–37
Hypertension 18 (21) 21 (10) 2.7 1.2–5.9
Immunologically mediated diseases
Asthma 2 (2.4) 8 (3.9) 0.7 0.1–3.1
Urticaria 12 (14) 29 (14) 1.0 0.5–2.0
Diabetes type I 0 (0) 0 (0) n.d.
Diabetes type II 0 (0) 5 (2.4) n.d.
Multiple sclerosis 0 (0) 2 (1.0) n.d.
Goitre 6 (7.1) 7 (3.4) 1.8 0.6–5.4
Inflammatory bowel disease 1 (1.2) 2 (1.0) 1.3 0.1–14
Gluten enteropathy 0 (0) 0 (0) n.d.
Psoriasis 2 (2.4) 8 (3.9) 0.6 0.1–2.9
Psoriatic arthropathy 0 (0) 2 (1.0) n.d.
Ankylosing spondylitis 0 (0) 1 (0.5) n.d.
Severe infections
Herpes zoster 5 (5.9) 12 (5.9) 1.0 0.4–2.9
Pneumonia 21 (25) 29 (14) 1.9 1.0–3.7
Meningitis 6 (7.1) 7 (3.4) 2.1 0.6–6.6
Osteitis 0 (0) 2 (1.0) n.d.
Pyelonephritis 8 (9.4) 15 (7.3) 1.3 0.5–3.2
Surgery
Major surgery 11 (13) 20 (10) 1.4 0.6–3.1
Blood transfusion 18 (21) 27 (13) 1.8 0.9–3.6
Others
Cancer 2 (2.4) 12 (5.9) 0.3 0.1–1.5
Gastric ulcer 6 (7.1) 4 (2.0) 3.1 0.9–11
a
Women without the disease in question were used as the reference category.
n.d., not determined.
TABLE 5. Distribution of cases and controls according to reported life events with corresponding OR and 95% CI
Cases Controls
Life event n (%) n (%) OR
a
95% CI
Death and disease
Severe disease of husband 9 (11) 22 (11) 1.0 0.4–2.9
Death of husband 6 (7.1) 14 (6.9) 1.0 0.3–3.2
Severe disease of child 8 (9.6) 14 (7.0) 1.3 0.5–3.3
Death of child 6 (7.1) 7 (3.4) 2.2 0.7–6.8
Economic stress
Dismissal 1 (1.2) 8 (3.9) 0.2 0.0–2.2
Unemployment 3 (3.7) 13 (6.4) 0.6 0.1–2.1
Severe economic problems 7 (8.5) 16 (7.9) 1.1 0.4–2.9
Conflicts
Divorce 14 (16) 29 (14) 1.3 0.6–2.8
Burglary or being robbed 6 (7.2) 31 (16) 0.4 0.2–1.0
Problems with police or judicial system 0 (0) 0 (0) n.d.
Severe conflict with someone 7 (8.3) 12 (5.9) 1.8 0.6–5.3
Deeply offended by someone 9 (11) 11 (5.5) 2.4 0.9–6.4
Others
Own severe disease (other than SLE) 12 (14) 29 (14) 0.9 0.4–2.0
Own serious accident 6 (7.1) 8 (3.9) 1.9 0.6–5.8
a
Women without the life event in question were used as the reference category.
n.d., not determined.
567Risk factors for developing SLE
Life events
The proportions of women with negative life events
and the corresponding ORs are summarized in Table 5.
For most life events included in the questionnaire,
no increased risk of SLE was observed. However, death
of a child and having been deeply offended were
reported more frequently by cases than by controls,
although the increased ORs were not statistically
significant.
Other items (hair colouring, animals and
alfalfa sprouts)
The OR was 1.6 (95% CI 0.8–2.9) for hair colouring
04 timesuyr compared with less frequent hair colouring,
but no tendency to a dose–response relationship was
seen (data not shown). The proportion of women who
reported consumption of alfalfa sprouts was similar
among cases and controls (12% and 15% respectively).
The proportions with close contact with dogs were
63% of the cases and 59% of the controls, with cats
47 and 43% respectively, with horses 31 and 28%,
with sheep 16 and 7.3%, and with goats 7.1 and 4.4%.
There was a significant difference between cases
and controls only for contacts with sheep (OR = 2.2,
95% CI 1.0–4.8).
Multivariate analyses
The associations between SLE and BMI, pneumonia,
cancer, gastric ulcer, hair colouring, exposure to sheep,
death of a child, having been deeply offended and a
family history of autoimmune diseases except SLE,
seen in univariate analyses, were not as obvious when
simultaneous adjustment was made for alcohol
consumption, smoking habits and hypertension in
multivariate analyses. On the other hand, a family
history of SLE, the sun-reactive skin type IuII and
drug allergy remained significantly associated with SLE
in the multivariate analyses. Moreover, blood transfu-
sion tended to be associated with SLE. Thus, the final
model included hypertension, drug allergy, a family
history of SLE, sun-reactive skin type IuII, blood
transfusion, smoking habits and alcohol consumption
(Table 6). The first four of these independent variables
were associated with a significantly increased risk of
developing SLE, while smoking habits and blood
transfusions were barely significantly associated with
SLE. Alcohol consumption was inversely associated
with the risk of SLE.
Discussion
The current view of the aetiology of SLE is that
several environmental factors act on the genetically
predisposed individual to either induce or protect from
disease w1x. The results of the present study suggest
roles for hypertension, drug allergy, skin type and
family history of SLE as risk factors for developing
SLE, and that alcohol is a potential protective factor.
The study was based on a population-based consecutive
case series of clinically well-characterized SLE patients,
which improves the relevance and validity of the
results. Furthermore, the response rate among cases
was high (93%).
Our data suggest that smoking is associated with
an increased risk of SLE, although this did not reach
statistical significance. However, no dose–response
relationship was observed. Our findings are in agree-
ment with results published previously w7, 8x. Alcohol
use has been suggested to be a protective factor. The
present investigation demonstrated a dose-dependent
inverse association between moderate alcohol use
and SLE, which was even more pronounced in the
multivariate model, further strengthening the observa-
tion. Our results are thus in agreement with the
two previous studies that have addressed this issue
specifically. However, in the study by Nagata et al.
w7x, the authors did not draw any firm conclusions
because of the infrequent alcohol intake in Japan,
and Hardy et al. w8x pointed out that their data
were derived from exposure after diagnosis and
therefore could be influenced by post-diagnosis
changes in alcohol consumption. Our data could be
influenced by recall bias, but the concordance of
the results from all three studies taken together
indicate strongly that the protective effect of alcohol
is real and that smoking increases the risk of
developing SLE.
The presence of aromatic amines in tobacco has
been suggested as an explanation of the link between
smoking and lupus, because drugs containing aro-
matic amines, such as hydralazine, are known inducers
of SLE w29x. Aromatic amines are also present in
hair-colouring products; one previous investigation
suggested an association between the use of these
TABLE 6. Variables included in the final multivariate model, with
corresponding adjusted OR and 95% CI
Variable Adjusted OR 95% CI
Alcohol consumption (gumonth)
0 1.0
1–150 0.4 0.2–1.0
>150 0.2 0.1–0.5
Smoking
Non-smoker 1.0
Smoker 1.8 0.9–3.6
Hypertension
No 1.0
Yes 3.7 1.4–9.8
Family history of SLE (first-degree relative)
No 1.0
Yes 6.8 1.4–32
Blood transfusion
No 1.0
Yes 2.3 0.9–5.8
Drug allergy
No 1.0
Yes 3.6 1.4–9.5
Skin type
III–IV (sometimesunever burn, always tan) 1.0
I–II (always burn, neverusometimes tan) 2.3 1.1–4.8
568 A. A. Bengtsson et al.
products and connective tissue disease w3x but later,
larger studies found less evidence of a link to SLE w4–6x.
We could not find any indications of an association
between hair colouring and SLE.
Among the self-reported diseases investigated in
this study, hypertension was associated with the most
pronounced increased risk of SLE. It is possible, though,
that vascular damage and endothelial cell dysfunction
contributing to hypertension in established SLE disease
w30x could be an early event, preceding clinical SLE
diagnosis, thus being a disease manifestation and not a
risk factor. On the other hand, endothelial cell dysfunc-
tion has been shown to predispose to immune complex
disease w31x. In addition, a history of hypertension is
nearly always combined with the use of anti-hyperten-
sive drugs, and among these are several that have been
associated with SLE and SLE-like drug reactions w32x.
The importance of drugs as possible SLE inducers
w32, 33x is illustrated further by our finding that a
history of drug allergy increased the risk of SLE
markedly. This finding is in agreement with the results
of a similar study, also investigating multiple risk
factors, by Strom et al. w16x. These authors found that
a history of allergic reactions or herpes zoster was
associated with increased risk of SLE. In our study,
most of the immunologically mediated diseases, includ-
ing allergy, were not associated with increased risk
of SLE. Regarding infectious diseases, a tendency
to increased risk of SLE was seen with pneumonia,
but in the multivariate model this increase in risk lost
its significance. Blood transfusion, which could repre-
sent the transmission of infectious agents, was
barely significantly associated with increased risk of
developing SLE.
The deleterious effect of UVR in SLE patients is
well known, and exacerbations of skin symptoms
during spring and summer have been described w34x,
as have UVR-induced systemic flares w35x. UVR
exposure as a risk factor for developing SLE has
been less investigated. We found that individuals who
constitutionally had a particular sun-reactive skin
type had an increased risk of developing SLE. These
individuals obviously tend to avoid extensive exposure
to the sun and as a consequence the cases were less
exposed to UVR compared with the controls in terms
of frequency of sunbathing, the use of a solarium and
going abroad to sunbathe (data not shown). Thus, our
results indicate that host factors, such as a particular
skin type, are the most important risk factors, and we
did not find any indication of excessive UVR exposure
prior to onset of SLE.
The results of the present study do not support
several postulated aetiological factors in SLE, including
exposure to hair-colouring products, dogs and exogen-
ous oestrogen. This could be because too few subjects
were investigated, which is the main drawback of this
study. Furthermore, the recruitment rate was consider-
ably higher in cases (93%) than in controls (53%), and
there were some possible indications of selection bias
in controls. The responders may have decided to
participate because of a history of some kind of
chronic disease or a family history of such a disease.
For example, unexpectedly many of the controls (9.9%)
reported that they had a first-degree relative with
rheumatoid arthritis. Furthermore, cancer was reported
more frequently in the control group than among the
cases. This prompted us to compare responders with
non-responders among the controls in the National
Cancer Register, in which all incident cancer cases
in Sweden are registered. In the responding control
group, comprising 205 subjects, the observed number
of cases with cancer was 19 and the expected number
of cancers in this group was 19.5. In contrast, the
observed number of individuals with cancer in the
non-responding group was 13 and the expected number
was 22.5.
Another source of error could be recall bias. Cases
would probably be more likely to make an effort when
filling in the questionnaire, but on the other hand a
considerable fraction of SLE patients have cognitive
dysfunction w36x. In addition, defining disease onset
may be complicated in a disease such as SLE. We chose
to use the year of diagnosis of SLE as this is the most
reliable time-point, but we are aware of the possibility
that this might have introduced a bias. Thus, the
retrospective nature of the present and similar investiga-
tions clearly has disadvantages. Not much is known
about the intervals between environmental exposure
and the development of autoimmunity, the first clinical
symptoms and full-blown disease. This leaves us uncer-
tain regarding the time relationship between exposure
and disease onset.
Notably, we did not find any indications that
hormonal factors play any role as risk factors for
SLE. Previously, only investigations on a larger number
of subjects have demonstrated that exposure to exogen-
ous oestrogen is associated with an approximately
doubled risk of SLE w12, 14–16x.
The interesting issue of an infectious agent causing
SLE being transmitted from animals, especially dogs,
to humans has been reviewed recently w19x. However,
we did not find any indications of a link between
animal exposure and the risk of SLE, although a
tendency was seen for exposure to sheep. Negative life
events as risk factors were also investigated without
finding any evidence of a connection.
As expected, the most obvious risk factor was a
first-degree relative with SLE, which was associated
with a more than six-fold increased risk of SLE. This
illustrates the importance of genetic factors and
suggests that environmental data and genetic data
should, if possible, be combined in future studies.
Thus, a specific environmental exposure might be a
risk factor only for the genetically predisposed indivi-
dual. In addition, when designing future studies
of specific exposures, such as hormonal factors, it
might be more informative to study certain age
groups or other selected groups of patients. On the
other hand, the methods used in the present investiga-
tion are well suited to the study of common exposures
569Risk factors for developing SLE
such as smoking habits and alcohol consumption.
When using a population-based, consecutive case
series of clinically well-characterized SLE patients,
risk factors of both exogenous and endogenous
origin can be identified, and further research in this
area is merited.
Acknowledgements
We gratefully acknowledge the help of Marita
Lindgren for collecting the data, Agneta Svedberg,
Cecilia Hemert and Zoli Mikoczy for designing
the questionnaire, Ralph Rittner for the selection of
controls and Palle Orbaek for advice on life event
questions. This work was supported by grants from
the Swedish Medical Research Council (Project
No. 13489), the Medical Faculty of the University of
Lund, the Swedish Rheumatism Association, the
80 Years Foundation of King Gustav V, the Greta
and Johan Kock Foundation, the Anna-Greta Craaford
Foundation, the Nanna Svartz Fund and the O
¨
sterlund
Foundation.
References
1. Cooper GS, Dooley MA, Treadwell EL, St Clair EW,
Parks CG, Gilkeson GS. Hormonal, environmental and
infectious risk factors for developing systemic lupus
erythematosus. Arthritis Rheum 1998;41:1714–24.
2. Reidenberg MM. Aromatic amines and the patho-
genesis of lupus erythematosus. Am J Med 1983;75:
1037–42.
3. Freni-Titulaer LW, Kelley DB, Grow AG, McKinley TW,
Arnett FC, Hochberg MC. Connective tissue disease in
southeastern Georgia: a case–control study of etiologic
factors. Am J Epidemiol 1989;130:404–9.
4. Petri M, Allbritton J. Hair product use in systemic lupus
erythematosus. A case–control study. Arthritis Rheum
1992;35:625–9.
5. Hardy CJ, Palmer BP, Muir KR, Powell RJ. Systemic
lupus erythematosus (SLE) and hair treatment: a large
community based case–control study. Lupus 1999;
8:541–4.
6. Sanchez-Guerrero J, Karlson EW, Colditz GA, Hunter DJ,
Speizer FE, Liang MH. Hair dye use and the risk of
developing systemic lupus erythematosus. Arthritis Rheum
1996;39:657–62.
7. Nagata C, Fujita S, Iwata H et al. Systemic lupus
erythematosus: a case–control epidemiologic study in
Japan. Int J Dermatol 1995;34:333–7.
8. Hardy CJ, Palmer BP, Muir KR, Sutton AJ, Powell RJ.
Smoking history, alcohol consumption and systemic lupus
erythematosus: a case–control study. Ann Rheum Dis
1998;57:451–5.
9. McAlindon T, Felson D, Palmer J, Zhang Y, Rosenberg L.
Associations of cigarette smoking and alcohol with
systemic lupus erythematosus (SLE) among participants
in the Black Womens Health Study (BWHS) wabstractx.
Arthritis Rheum 1997;40:162.
10. McAlindon T, Formica M, Palmer J, Rosenberg L. Is
childhood smoke exposure a riskfactor for SLE? Results
from the Black Women’s Health Study (BWHS) wabstractx.
Arthritis Rheum 2000;43:237.
11. Reidenberg MM, Drayer DE, Lorenzo B et al. Acetylation
phenotypes and environmental chemical exposure of
people with idiopathic systemic lupus erythematosus.
Arthritis Rheum 1993;36:971–3.
12. Sanchez-Guerrero J, Liang MH, Karlson EW, Hunter DJ,
Colditz GA. Postmenopausal estrogen therapy and the
risk for developing systemic lupus erythematosus. Ann
Intern Med 1995;122:430–3.
13. Meier CR, Sturkenboom MC, Cohen AS, Jick H.
Postmenopausal estrogen replacement therapy and the
risk of developing systemic lupus erythematosus or discoid
lupus. J Rheumatol 1998;25:1515–9.
14. Sanchez-Guerrero J, Karlson EW, Liang MH, Hunter DJ,
Speizer FE, Colditz GA. Past use of oral contraceptives
and the risk of developing systemic lupus erythematosus.
Arthritis Rheum 1997;40:804–8.
15. Grimes DA, LeBolt SA, Grimes KR, Wingo PA.
Systemic lupus erythematosus and reproductive function:
a case–control study. Am J Obstet Gynecol 1985;
153:179–86.
16. Strom BL, Reidenberg MM, West S, Snyder ES,
Freundlich B, Stolley PD. Shingles, allergies, family
medical history, oral contraceptives and other potential
risk factors for systemic lupus erythematosus. Am J
Epidemiol 1994;140:632–42.
17. Schwartz RS. Viruses and systemic lupus erythematosus.
N Engl J Med 1975;293:132–6.
18. Jones DR, Hopkinson ND, Powell RJ. Autoantibodies in
pet dogs owned by patients with systemic lupus erythe-
matosus. Lancet 1992;339:1378–80.
19. Panush RS, Levine ML, Reichlin M. Do I need an ANA?
Some thoughts about man’s best friend and the transmis-
sibility of lupus. J Rheumatol 2000;27:287–91.
20. Winsa B, Adami HO, Bergstrom R et al.
Stressful life events and Graves’ disease. Lancet 1991;
338:1475–9.
21. Thernlund GM, Dahlquist G, Hansson K et al.
Psychological stress and the onset of IDDM in children.
Diabetes Care 1995;18:1323–9.
22. Wallace DJ. Does stress or trauma cause or aggravate
rheumatic disease? Bailliere’s Clin Rheumatol 1994;
8:149–59.
23. Sontheimer RD. Photoimmunology of lupus erythemato-
sus and dermatomyositis: a speculative review. Photochem
Photobiol 1996;63:583–94.
24. Malinow MR, Bardana EJ Jr, Goodnight SH Jr.
Pancytopenia during ingestion of alfalfa seeds. Lancet
1981;1:615.
25. Jonsson H, Nived O, Sturfelt G, Silman A. Estimating the
incidence of systemic lupus erythematosus in a defined
population using multiple sources of retrieval. Br J
Rheumatol 1990;29:185–8.
26. Jonsson H, Nived O, Sturfelt G. Outcome in systemic
lupus erythematosus: a prospective study of patients from
a defined population. Medicine (Baltimore) 1989;
68:141–50.
27. Fitzpatrick TB. The validity and practicality of sun-
reactive skin types I through VI. Arch Dermatol
1988;124:869–71.
28. Hosmer DW, Lemeshow S. Applied logistic regression.
New York: John Wiley and Sons, 1989.
29. Reidenberg MM, Durant PJ, Harris RA, De Boccardo G,
Lahita R, Stenzel KH. Lupus erythematosus-like disease
due to hydrazine. Am J Med 1983;75:365–70.
570 A. A. Bengtsson et al.
30. Manzi S. Systemic lupus erythematosus: a model for
atherogenesis? Rheumatology 2000;39:353–9.
31. Geirsson AJ, Sturfelt G, Truedsson L. Clinical and sero-
logical features of severe vasculitis in rheumatoid arthritis:
prognostic implications. Ann Rheum Dis 1987;46:727–33.
32. Adams LE, Hess EV. Drug-related lupus. Incidence,
mechanisms and clinical implications. Drug Safety 1991;
6:431–49.
33. Fritzler MJ. Drugs recently associated with lupus syn-
dromes. Lupus 1994;3:455–9.
34. Sturfelt G, Nived O. Clinical inconsistency, benign course
and normal employment rates in unselected systemic lupus
erythematosus. Clin Exp Rheumatol 1985;3:303–10.
35. Nived O, Sturfelt G, Ryberg B. Two UV-induced episodes
of myelitis in a patient with systemic lupus erythematosus.
J Intern Med 1992;232:461–3.
36. Ginsburg KS, Wright EA, Larson MG et al. A controlled
study of the prevalence of cognitive dysfunction in
randomly selected patients with systemic lupus erythema-
tosus. Arthritis Rheum 1992;35:776–82.
571Risk factors for developing SLE
