Prognosis in systemic lupus erythematosus.

37
PROGNOSIS IN SYSTEMIC LUPUS ERYTHEMATOSUS
Negative Impact of Increasing Age at Onset, Black Race, and
Thrombocytopenia, as well as Causes of Death
JOHN D. REVEILLE, ALFRED BARTOLUCCI, and GRACIELA S. ALARCdN
To assess the impact of demographic and clinical
factors on prognosis in patients with systemic lupus
erythematosus (SLE), we examined survivorship by
Life-table analysis in 389 patients. There were approximately equal numbers of Caucasian patients and American black patients in this study group. On both univariate and multivariate analyses, we found that both
American black race and increasing age at SLE onset
independently worsened the probability of survival. Of
all the clinical factors we analyzed, thrombocytopenia
emerged as the only independent risk factor for a worse
prognosis in SLE. In all clinical and demographic
groups considered, the leading cause of death was
infection.
Systemic lupus erythematosus (SLE) is a
chronic, multisystem autoimmune disease that predominantly affects women of childbearing age. The
extensive cutaneous, musculoskeletal, and visceral
manifestations of this disease, especially in younger
patients, are often associated with considerable limiFrom The Departments of Medicine (Division of Clinical
Immunology and Rheumatology and the Clinical Information Systems), Biostatistics and Biomathematics, and Epidemiology, School
of Medicine and School of Public Health, The University of Alabama at Birmingham.
John D. Reveille, MD: Department of Medicine (current
address: Division of Rheumatology and Clinical Immunogenetics,
University of Texas Health Science Center, Houston); Alfred Bartolucci, PhD: Department of Biostatistics and Biomathematics;
Graciela S. Alarcon, MD, MPH: Department of Medicine.
Address reprint requests to John D. Reveille, MD, Division
of Rheumatology and Clinical Immunogenetics, University of Texas
Health Science Center, PO Box 20708, Houston, TX 77225.
Submitted for publication December 20, 1988; accepted in
revised form August 7, 1989.
Arthritis and Rheumatism, Vol. 33, No. 1 (January 1990)
tation of activities, both social and occupational, as
well as with a decreased life expectancy.
Since the initial, 1955 report of a markedly
decreased probability of survival of patients with SLE
(l), prognosis in this disease has been the subject of a
number of studies (2-16). The results of these studies
show that the probability of survival has significantly
improved over the ensuing 35 years, although it is
debatable whether this is due to increased awareness
and earlier diagnosis in less severely affected individuals, or to better treatment. Aside from the negative
impact of renal and neuropsychiatric manifestations
on outcome in SLE (10,l l), few studies have identified
a role for other clinical features of the disease in the
prognosis. Demographic characteristics, particularly
race (5,9,15-18), sex (19), and young age at disease
onset (17,18,20-26), have also been shown to affect the
clinical presentation of SLE. Moreover, this disease
has been shown to be more common among the
American black population (9,27), and some studies
have found a worse prognosis for American black
patients with SLE than for Caucasian SLE patients
(5,15,16). However, the reasons for this difference,
whether due to socioeconomic or clinical factors (or a
combination of these), are not clear.
To further address these issues, we studied a
large, geographically stable, and clinically well-defined
group of SLE patients to determine the relative contributions of demdgraphic and clinical factors to disease outcome, and to causes of death. The results of
our epidemiologic analyses are reported here.
PATIENTS AND METHODS
Patients in this study were seen by members of the
Division of Clinical Immunology and Rheumatology or the
38
Division of Nephrology at The University of Alabama at
Birmingham (UAB) affiliated hospitals, as either outpatients
or inpatients, between 1975 and 1984. This included patients
seen at the UAB hospital and outpatient clinic, the Birmingham Veterans Administration Hospital, and the Birmingham
Children’s Hospital.
International Classification of Disease codes 734. I
(before 1978) and 710.0 (after 1978) were used to perform a
computer search for all inpatients seen during the period of
study. In addition, all medical records from the rheumatology outpatient clinic (covering the period 1979 to 1984) were
reviewed for those with the diagnosis of SLE who were seen
as outpatients only. Outpatients seen by members of the
Division of Clinical Immunology and Rheumatology between 1975 and 1979 were identified by reviewing laboratory
serum log books, since sera from nearly all SLE outpatients
during this period had been stored. Although outpatient
records from the Division of Nephrology and the Birmingham Children’s Hospital were not reviewed, the names of all
those with SLE who were seen as outpatients between July
1983 and July 1985 were provided by their respective physicians. The charts of all patients seen in the rheumatology
outpatient clinic at the VA Hospital between July 1982 and
July 1985 were also reviewed for the diagnosis of SLE (lists
of patients seen at this facility before July 1982 were not
available).
Only those patients whose SLE met 4 or more of the
American Rheumatism Association (ARA) revised criteria
for the classification of SLE (28) were included in this study.
Patients referred for renal transplant or for hemodialysis
alone were studied only if their referring physician had
included records of their disease course.
The medical records of each patient were abstracted
for the following information: race, sex, date of birth, age at
SLE onset, date of onset of SLE (defined as the onset of
multisystem disease), insurance status at the first visit to our
medical system (private insurance versus public insurance,
which was defined as Medicare a n d o r Medicaid only, or no
insurance), and absence or presence (at any time in the SLE
course) of the following disease features: arthritis, neuropsychiatric manifestations (including seizures, psychosis, longtract signs, coma, transverse myelitis, cranial nerve signs,
chorea, organic brain syndromes, or other central nervous
system [CNS] lesions directly attributable to SLE and not to
metabolic, drug, or other causes, and excluding headache
and peripheral neuropathy), renal disease (defined as 1 or
more of the following: 3 0 0 mg of proteinuria in 24 hours,
>2+ proteinuria on 2 occasions, or cellular casts [not
attributable to infection], or abnormal renal biopsy findings),
mucocutaneous disease (and type), serositis (including pericarditis, as documented by echocardiogram, or pleurisy, as
documented by the presence of pleuritic chest pain or pleural
effusion believed by the physician of record to be due to SLE
and not to infection), or cytopenia (either as autoimmune
hemolytic anemia, leukopenia of 54,000 white blood cells
[WBC]/mm3, or moderate-to-severe thrombocytopenia defined as 5100,000 platelets/mm3). The maximum level of
anti-DNA antibodies that had been detected was entered and
subcategorized as high titer (250% binding by the Farr
technique [29] or 2 1:80 by Crithidia luciliae [30]), low titer
REVEILLE ET AL
(10-50% binding on Farr assay or <1:80 on C luciliae
substrate), or absent.
To analyze the impact of clinical parameters of SLE
present at the time of, or within 6 months of, the diagnosis,
we further examined a subset of patients whose medical care
from the time of diagnosis had been at our system or whose
complete medical records since diagnosis were available for
our review. The following data were obtained at entry into
the medical system for those patients seen at, or within 6
months of, the SLE diagnosis: age at diagnosis, date of
diagnosis, serum creatinine level (elevated levels defined as
2 1.5 mg/dl), amount of proteinuria, blood pressure value,
hematocrit value, platelet count (moderate-to-severe thrombocytopenia defined as 5 100,000/mm3), DNA antibody levels, and serum hemolytic complement (CH50) levels (hypocomplementemia defined as depressed CH50), as well as the
presence of neuropsychiatric manifestations, myositis, and
pulmonary involvement (defined as a chest radiograph showing interstitial fibrosis or pulmonary function tests demonstrating restrictive lung disease).
Causes of death were determined on the basis of the
clinical data, death certificates, and when available, on
autopsy findings. For SLE patients who died at the UAB
affiliated hospitals, the following laboratory values were
obtained from the medical records at the time of the last
hospital admission: blood glucose, serum creatinine, hematocrit, WBC count, differential cell count, and platelet count.
We also recorded the dosage of prednisone the patient had
been taking prior to admission, as well as any use of
immunosuppressive agents, either at or after the last hospital
admission.
In the fall of 1985, we contacted all of the patients,
their family members, or their referring physicians if the
family members were not available, to ascertain the patient’s
current status. Death certificates were sought from the
Alabama Bureau of Vital Statistics for those patients who
could not be located. The duration of followup was defined
as the interval from onset of multisystem disease. In the
subset of patients with medical records available from the
time of SLE diagnosis, the duration of followup began at the
time the diagnosis of SLE was established. For those whose
current status could not be ascertained, followup was calculated as the interval from the onset or diagnosis of SLE to
the last contact with the system.
Life-table analysis was performed by the method of
Merrell and Shulman ( I ) . The method proposed by Greenwood (31) was used to determine difFerences between survival curves. We also utilized Cox’s proportional hazards
general linear model procedure (32), with the dependent
variable being survival, to determine which of the many
potentially important parameters had the greatest impact on
prognosis in SLE. For life-table and multivariate analyses, P
values less than or equal to 0.05 were considered significant.
RESULTS
Demographic characteristics of patients studied.
A total of 525 medical records with a listed diagnosis of
SLE were reviewed. Of these, 67 patients were found
39
FACTORS AFFECTING SLE PROGNOSIS
Table 1. Impact of demographic and clinical parameters, present at any time during the course of
systemic lupus erythematosus, on survival*
Probability of survival (%)
n
year
5
years
years
15
years
85
304
96.0
96.0
89.1
88.8
82.8
83.8
80.2
79.0
NS
151
97
95.2
93.5
92.5
84.7
86.2
80.2
-
<0.01
33
106
95.5
96.1
90.0
78.9
81.9
74.8
-
NS
215
174
95.3
95.4
86.5
91.2
78.5
89.4
71.6
87.6
<0.01
68
86
95.6
92.9
83.3
82.3
81.7
72.7
-
<0.05
101
288
90.2
97.2
83.0
89.1
74.8
84.7
71’7
80.7
<0.05
79
309
85.9
97.7
74.0
90.6
65.2
85.3
81.7
c0.005
1
Males
Females
Insurance statust
Private insurance
Caucasians
American blacks
Public o r no insurance
Caucasians
American blacks
Renal disease
Present
Absent
ProteinuriaS
0.5-3 g d 2 4 hours
>3 g d 2 4 hours
Neuropsychiatric disease
Present
Absent
Thromboc ytopenia8
Present (5100,000 plateletslmm’)
Absent
10
P
C0.03
* Total patient population was 389. Time interval from the onset of multisystem autoimmune disease
to either fall 1985, death, or last followup. NS = not significant; - = number too small for meaningful
analysis.
t Insurance status was not available for 2 patients; the 12 patients from the VA hospital were included
in the group with public or no insurance.
$ Compared with patients without proteinuria (<0.5 g d 2 4 hours). No significant difference was seen
in the probability of survival when the survival curve of those with 0.5-3 gm of protein in 24 hours was
compared with those with >3 grn/24 hours. No 24-hour urine protein collections were available from
61 patients with renal involvement.
8 Platelet count was not available for I patient.
to have diagnoses other than SLE (including 10 with
drug-induced lupus), 14 had been referred for renal
transplant evaluation or hemodialysis but without the
records of their disease course, and 15 had incomplete
data available to evaluate the SLE diagnosis. Forty
patients had a listed diagnosis of SLE and complete
clinical and serologic data, but their symptoms did not
meet 4 of the ARA revised criteria for the classification
of SLE; they were therefore not included in this study
(19 were American black and 21 were Caucasian).
In all, 389 patients who met 4 or more ARA
revised criteria for the classification of SLE and were
seen at UAB between 1975 and 1984 (including 12
referred for hemodialysis or transplant alone and I2
who were seen only at the VA hospital) were identified. Of these, 184 were Caucasian, 203 were American black, and 2 were Asian. There were 85 male and
304 female patients. Caucasian males were significantly older at SLE onset than were Caucasian fe-
$
1
P< 0.05
3
5
0
‘0
2
4
6
8
10
12
14
33
38
25
Years Since Disease Onset
Caucasians-184
AmericanBlacks ....2 0 3
149
161
121
125
96
86
70
66
47
53
32
Figure 1. Cumulative probability of survival at disease onset for
387 patients with systemic lupus erythematosus, according to race
(184 Caucasians and 203 American blacks; 2 Asians excluded).
Differences between survival curves were determined by the
method of Greenwood (31).
REVEILLE ET AL
40
*
males (mean ? SD age 47.8 +- 17.6years versus 36.8
16.7 years; P < 0.001). Caucasian males and females
as a group were significantly older at disease onset
than were American black females (27.2 13.1 years)
(P < 0.001) or males (27.6 17.5 years) (P < 0.001).
There was an increased frequency of SLE in females
compared with males, which was much more significant among the American blacks than among the
Caucasians (ratio of fema1es:males 6.1: 1 in American
blacks versus 2.2:l in Caucasians; P < 0.001). The
mean duration of followup was 8.8 years. The fall 1985
*
*
status was obtained for 363 of the original 389 patients
(93%), including 6 whose death certificates were obtained from the Alabama Bureau of Vital Statistics. A
total of 89 patients (25%) had died.
Impact of demographic parameters on survival.
Univariate analysis demonstrated no difference in the
probability of survival of males compared with females
with SLE (Table 1). When race alone was considered,
however, American black SLE patients fared worse
than Caucasian SLE patients (P < 0.05) (Figure 1).
Those patients with SLE onset before the age of 20
n
F
Y
.02
r
0
t
6o
40.
.-
20.
-a
.I-
3
5
2
0
4
6
10
8
Years Since Disease Onset
Years Since Disease Onset
SLE Patients-94
U.S. Population
----
85
66
49
40
SLEPatients-218
30
202
162
135
104
84
US. Population.---
B
A
h
z#
Y
100-
Figure 2. Cumulative probability of survival at disease onset for
389 patients with systemic lupus erythematosus (SLE), according to
age at disease onset, as compared with the US population. A. Age
0-19 years (94 patients). B. Age 2 0 4 9 years (218 patients). C. Age
2 5 0 years (77 patients). Differences between survival curves were
determined by the method of Greenwood (31).
Years Since Disease Onset
SLEPatients-77
52
68
US. Population----
C
35
21
15
41
FACTORS AFFECTING SLE PROGNOSIS
appeared to do worse than those with later onset
(between ages 20 and 49), and the probability of
survival in the group with young age at SLE onset
approximated that of the group with onset at age 50 or
after.
Compared with the estimated survival of the
age-matched segment of the US population (33) (data
for age-matched segment of the Alabama population
not available), SLE patients fared significantly worse
in all age groups (Figures 2A-C). The most striking
difference was in the group whose SLE began when
they were young. However, when compared with each
other, the differences in probability of survival between the 3 age groups considered did not reach
statistical significance.
Insurance status (as a proxy for economic status) indicated a slight difference in the probability of
survival between those with private insurance compared with those with public insurance or no insurance,
although these differences did not achieve statistical
significance. Moreover, Caucasian SLE patients with
private insurance fared better than American black
SLE patients with private insurance (P < 0.01). The
number of Caucasian patients who had public insurance or no insurance (n = 33) was too small to perform
conclusive life-table analysis. When American black
patients with private insurance (n = 97) were compared with those with public insurance or no insurance
(n = lM), the differences did not reach statistical
significance.
Impact of clinical parameters, occurring at any
time during the course of SLE, on survival. Because
arthritis, serositis, autoimmune hemolytic anemia, leukopenia, and cutaneous disease were equally present
in those who died and those who did not die (data not
shown), life-table analysis was not performed for those
parameters. Patients with renal involvement at any
time in the disease course (n = 215) fared worse than
those with no renal involvement (n = 174) (P< 0.01).
Furthermore, the maximum level of proteinuria at any
time in the disease course (>3 gm/24 hours and 0.5-3
g d 2 4 hours versus <0.5 g d 2 4 hours) also adversely
affected prognosis (Table 1).
Similarly, neuropsychiatric manifestations negatively impacted survival probability (P < 0.05). Because there were only 13 patients with only psychiatric
symptoms, statistical analysis of this group versus
those with neurolokic involvement would not be meaningful. However, the most striking negative effect on
survival, more than either renal or neuropsychiatric
involvement, was qignificant thrombocytopenia (platelet count 5100,000!mm3, n = 79) (Figure 3), with only
-
100
c
0
0
Thrombocylopenia
Presenl
Absent
____
-
2
4
6
8
10
12
14
18
12
58
12
45
8
30
Years Since Disease Onset
78
61
309
253
50
109
30
156
23
116
17
82
Figure 3. Cumulative probability of survival at disease onset for 79
systemic lupus erythematosus (SLE) patients who had thrombocytopenia at any time in the disease course, compared with 309 SLE
patients without thrombocytopenia (1 patient in whom platelet
counts were not available was excluded). Differences between
survival curves were determined by the method of Greenwood (31).
an 85.9% probability of survival 1 year after disease
onset (P < 0.005). After 2 years, the slope of the
survival curve leveled off, and paralleled the slopes of
the renal and neuropsychiatric involvement curves.
This suggests that the major impact of moderateto-severe thrombocytopenia on SLE prognosis occurs
in the first 2 years of disease.
Multivariate analysis showed that American
black race, increasing age at SLE onset, and moderate-to-severe thrombocytopenia, were independent
factors that adversely affected survival in SLE (Table
2). When adjusted for other variables in the model,
including sex, insurance status, neuropsychiatric involvement, DNA antibodies, proteinuria, renal disease, and serositis, none of the clinical parameters
considered (present at any time in the course of SLE)
affected the probability of survival.
Table 2. Cox multivariate analysis of the impact of demographic
and clinical parameters occumng at any time during the course of
systemic lupus erythematosus on survival*
Variable
Increasing age at disease
onset
American black race
Thrombocytopenia
J
P
12.33
O.OOO4
9.93
4.42
0.0016
0.035
* Total patient population was 389. Only those parameters found to
have a statistically significant impact are shown. Insurance status,
neuropsychiatric involvement, presence of DNA antibody, proteinuria, renal involvement, and serositis showed no significant impact
on prognosis by multivariate analysis.
42
REVEILLE ET AL
Impact of clinical parameters, at diagnosis, on
survival. For the 277 patients on whom extensive
clinical data were available either at or within 6
months of SLE diagnosis, there were significant differences in the results of univariate life-table analysis
from the time of diagnosis (Table 3). When present at
the time of diagnosis, proteinuria, diastolic or systolic
hypertension, anemia, and thrombocytopenia all negatively impacted survival probability. Of the 79 patients who had a platelet count 5100,000/mm3 at any
time in the disease course, platelet counts done at or
within 6 months of SLE diagnosis were available in 61.
Qf these 61 patients, 37 (61%) had a platelet count
s100,000/mm3at diagnosis. It should be noted that for
the analysis shown in Table 3, thrombocytopenia was
defined as a platelet count of <150,000/mm3. The
probability of survival was the same for those with a
platelet count of I100,W/mm3(data not shown), but
fewer patients in this group meant less meaningful
statistical analysis.
Thrombocytopenia was present with approximately equal frequency in patients with and patients
without renal disease (23% versus 19%), although it
was slightly more frequent in patients with neuropsychiatric disease than in those without this manifestation (30% versus 19%; P = 0.02). The presence of
hypocomplementemia, elevated serum creatinine levels, DNA antibodies (either low or high titer), neuropsychiatric manifestations, and interstitial lung disease did not adversely affect prognosis in this
subgroup of patients. The negative influence of some
parameters, such as thrombocytopenia (even when
mild) and hypertension, were apparent early in the
disease course, whereas other parameters, such as
Table 3. Impact of clinical parameters, present at or within 6 months of diagnosis of systemic lupus
erythematosus, on survival*
Probability of survivial (%)
1
year
5
years
years
P
170
92
93
81
86
72
82
0.02
225
52
94
87
85
77
79
-
NS
60
217
88
94
80
85
81
0.04
n
Proteinuriat
20.5 g d 2 4 hours
<0.5 gm/24 hours
Serum creatinine
5 1 . 5 gm/dl
>1.5 gm/dl
Systolic blood pressure
>I50 mm Hg
5150 mm Hg
Diastolic blood pressure
>90 m m Hg
590 mm Hg
Hematocritt
<30%
2 30%
Platelet countt
<150,000/mm3
2150,000/mm3
Neuropsychiatric disease
Present
Absent
Interstitial lung disease
Present
Absent
DNA antibodiest
Absent
Present at low titer
Present at high titer
Hypocomplementemiat
Present
Absent
106
10
50
86
80
227
94
84
80
0.01
88
188
89
94
75
87
70
81
0.05
61
21 1
82
96
72
92
67
83
0.005
47
230
90
93
87
83
49
228
90
93
81
84
54
I08
40
98
94
93
95
86
83
80
151
82
91
93
79
88
74
80
NS
NS
* Total patient population was 277. Time interval from the diagnosis to either fall 1985, death, or last
followup. NS = not significant; - = number too small for meaningful analysis.
t Parameter was not measured at diagnosis in all patients.
FACTORS AFFECTING SLE PROGNOSIS
Table 4. Cox multivariate analysis of the impact of demographic
and clinical parameters present at or within 6 months of the
diagnosis of systemic lupus erythematosus on survival*
Variable
2
P
American black race
Thrornboc ytopenia
Increasing age at disease
onset
9.45
8.82
4.52
0.002
0.003
0.034
* Total patient population was 277. Only those parameters found to
have a statistically significant impact are shown. Hypocomplementemia, neuropsychiatric involvement, elevated diastolic or systolic
blood pressure, decreased hematwrit, presence of DNA antibody,
pulmonary disease, insurance status, elevated creatinine level, and
proteinuria showed no significant impact on prognosis by multivariate analysis.
proteinuria, appeared to produce their greatest effect
later.
Multivariate analysis of demographic and clinical factors present at or within 6 months of disease
diagnosis showed that American black race, thrombocytopenia (defined, as in the previous multivariate
analysis, as a platelet count %100,000/mm3),and increasing age at SLE onset were independent risk
factors for a decreased probability of survival (Table
4). Regardless of their effect on SLE outcome as
shown by univariate analysis, neither proteinuria, anemia, hypertension, hypocomplementemia, DNA antibodies, neuropsychiatric involvement, interstitial lung
disease, elevated creatinine level, nor insurance status
at diagnosis independently affected the prognosis in
SLE as shown by multivariate analysis.
Causes of death in the study population. Eightynine patients died between 1975 and the end of 1985
(26% of the American black SLE patients and 20% of
the Caucasian SLE patients studied). The circumstances and/or cause of death were known in 74 study
patients (83%). In this group, infection was the leading
primary cause of death (Table 5 ) , followed by bleeding
(gastrointestinal or CNS hemorrhage), cardiovascular
disease, and active SLE. The proportions of those
who died of infection compared with those who died of
other causes were equivalent in Caucasian patients
and American black patients, as well as in all patients
with and all patients without thrombocytopenia (data
not shown). Infection as a primary cause of death was
more frequent in the early-onset SLE group (46%, age
519) and in the “mid-onset’’ group (61%, ages 2049)
than in the group with late-onset SLE (26%, age ?SO)
(P= 0.03). In this latter group, cardiovascular disease
and neoplasms, respectively, accounted for 26% and
21% of the deaths, compared with 8% and 4% of the
43
Table 5. Causes of death of patients with systemic lupus erythematosus (SLE). according to age at disease onset*
Age at onset of SLE
5 1 9 years
(n = 24)
2 0 4 9 years
(n = 31)
2 5 0 years
Cause of death
Infection
Gram-positive
Gram-negative
Mixed bacterial
Fungal
Unspecified
Bleeding
Active SLE
Renal failure7
Neoplasm
Cardiovascular disease
Other
46t
4
21
4
4
13
13t
17
8
4
8
17
61$
16
26
0
13
7
10
1O$
3
0
3
20
265
0
5
5
0
16
5
11
0
21
26
11
(n
=
19)
* A total of 89 patients died during the period of study; cause of
death was ascertained for 74 of them. Values shown are percentages.
t Includes 2 patients who died of sepsis and bleeding, and 1 who
died of sepsis in the setting of uremia.
$ Includes 2 patients who died of infection in the setting of active
SLE.
§ P = 0.03 versus all patients under age 50 at SLE onset.
fl Includes those with complications of either uremia or dialysis.
deaths in the early-onset SLE group and 3% and 0%
in the mid-onset group (P = 0.0003, relative risk
[RR] = 13).
Of the 29 SLE patients with moderate-to-severe
thrombocytopenia ( I l00,00O/mm~at any time in the
disease course) whose cause of death was known, only
5 (17%) died from hemorrhage or complications
thereof, compared with 3 (7%) of the 45 patients
without thrombocytopenia whose cause of death was
known. Of the 28 patients with thrombocytopenia who
died (and for whom clinical data were available; in 1
patient, the cause of death was recorded, but no other
clinical data were available), 14% were taking immunosuppressive drugs in the month prior to death,
compared with 40% of patients without thrombocytopenia who died. Among the 50 patients whose prednisone dosage at the time of the last hospital admission
(before death) was known, the mean
SD dosage
varied widely, from 32.0 f 35.8 mg/day in the 23
patients with thrombocytopenia to 24.4 ? 25.0 mg/day
in the 27 patients without (Pnot significant). Similarly,
there were no significant differences in the mean
prednisone dosage at the time of the last hospital
admission in the 30 patients with SLE who died of
infection (28.5 25.4 mg/day) compared with the 20
who diqd of other causes (20.3 2 36.4 mg/day). Likewise, the frequency of recent immunosuppressive
*
*
REVEILLE ET AL
44
therapy did not differ between the 2 groups (24%
versus 17%). It is noteworthy that renal insufficiency
was frequently present on the first day of the final
hospital admission both in those who died of infection
(mean ? SD serum creatinine 3.5 ? 3.5 mg/dl) and in
those who died of other causes (5.0 4.9 mg/dl).
*
DISCUSSION
We have examined the impact of specific clinical and demographic characteristics on the probability
of survival in a well-defined group of 389 SLE patients.
The numbers of American black patients and Caucasian patients were equally well represented (n = 203
and n = 184, respectively). The parameters were
examined for their effects when present at SLE diagnosis, as well as when present at any time in the course
of the disease. Knowing what clinical features are
present at diagnosis is important to such analyses
because of their predictive value; moreover, it is less
likely that such features are the result of progressive
disease. For example, proteinuria could be the result
of longstanding hypertension, or the hypertension
could itself be the result of longstanding lupus nephritis. The disease features that emerge later in the course
of SLE also require analysis, because they may be
markers of more severe SLE and may require more
aggressive therapy. Since medical records at or within
6 months of diagnosis were not available for 112 SLE
patients, the parameters present at diagnosis were
analyzed in the smaller group of SLE patients whose
records were available (n = 277).
On both univariate and multivariate analyses,
we found that American black race acts as a risk factor
for a worse prognosis; that is, American black patients
with SLE have a lower life expectancy than Caucasian
patients with SLE, regardless of the other demographic and clinical features present. In a previously
published multicenter study of outcome in SLE (UAB
not included) (15)-an
analysis of 642 Caucasian patients and 356 black patients (including American and
Caribbean blacks )-this
difference in prognosis was
also noted, with black patients having more severe
disease at study entry (as evidenced by a higher mean
serum creatinine level, lower mean hematocrit value,
and higher frequency of CNS lupus). It should be
noted that, although we did not analyze place of birth
data for all subjects in this study, of the 100 black SLE
study patients currently followed by us, 92 were born
in Alabama, 6 were born elsewhere in the southeastern
US, and 4 were born elsewhere in the US. We are not
aware of any black SLE patients in this study who
were born outside the US.
Studenski et a1 (16), who studied SLE patients
seen at another Southern US medical center, found a
decreased probability of survival among nonwhites.
Other, earlier reports suggesting that Caucasians and
American blacks have a similar probability of survival
(4,6,13) are complicated by the fairly small numbers of
non-Caucasians studied. Thrombocytopenia, whose
presence at diagnosis was the only clinical feature that
was an independent risk factor for a poorer outcome,
was present in this study in equal frequency in Caucasians (30%) and American blacks (29%), except in
the group with late-onset SLE, in whom severe thrombocytopenia was slightly more common in American
blacks (40%) (data not shown).
Results of a multicenter study by Ginzler et a1
( 1 3 , as well as the recent study by Studenski et al(16),
suggested socioeconomic factors (using insurance status as a proxy) to explain at least part of the dif€erences in mortality rates between blacks and Caucasians. However, even in the multicenter study, the
correlation of a worse prognosis with public funding
varied from center to center. We found the probability
of survival for American blacks with private medical
insurance was not significantly better than that for
American blacks with public or no medical insurance.
Moreover, multivariate analysis did not demonstrate
insurance status as producing a negative impact on
prognosis, which contrasts with Studenski and coworkers’ findings (16). Since both groups were composed of American blacks from the Southern US,
there is no apparent reason for the discrepant findings.
Perhaps either insurance status does not correlate with
socioeconomic status, and therefore cannot be used to
represent this feature, or socioeconomic status does
not truly act as a significant factor in the diminished
probability of survival of American blacks compared
with Caucasians. A study examining socioeconomic
status in more detail, such as one using the “Hollingshead index” (34), which is widely used among social
and behavioral scientists, may help to further address
this issue by analyzing the impact of educational status
and occupation (among many other variables). Esdaile
et al (35) found no correlation between socioeconomic
status (as determined from educational and occupational status) and outcome (as measured not by survival but by health status, which was determined by
the Arthritis Impact Measurement Scales scores) in 78
Canadian SLE patients. On the other hand, Hochberg
and Sutton (36) reported that educational status was
FACTORS AFFECTING SLE PROGNOSIS
associated with poor psychosocial adjustment to SLE;
thus, it appears that at least some socioeconomic
factors may be important in the impact of SLE.
At least 2 interrelated factors are more likely to
be present in American blacks with SLE, and these
may contribute to their lowered expected survival.
These patients had a higher frequency of renal involvement (a%,
versus 46% in Caucasians; data not
shown), a finding that has been reported by others
(18), and more aggressive disease at onset (indicated
by the shorter mean interval between the diagnosis of
SLE and the onset of multisystem disease). It is
noteworthy that, despite the results of the univariate
analysis, neither the presence of renal disease nor the
presence of an elevated serum creatinine level at
diagnosis independently affected prognosis; this contrasts with the findings of the previous multicenter
study (15). Whether this is due to differences in
treatment or in compliance is not clear, and the finding
merits further study.
Sixty-one patients with renal involvement did
not have 24-hour protein excretion determinations.
These patients might represent a group with transient
or milder renal involvement. The possible effect this
might have on the data would be to falsely increase the
impact of proteinuria on prognosis in SLE, but multivariate analysis demonstrated no association of proteinuria with a worse prognosis. In patients who died
(regardless of cause), the elevated serum creatinine
level noted on the first day of the last hospital admission (before death),could have been due to a variety of
causes, including lupus nephritis, hypertension, heart
failure, or sepsis, or combinations thereof.
Age at SLE onset proved to be an important
determinant in prognosis. On univariate analysis, both
early-onset and late-onset SLE groups appeared to
fare worse, whereas on the multivariate analysis, only
increasing age at SLE onset independently worsened
the probability of survival. The higher proportion of
American blacks (this being another independent risk
factor) in the young age at onset group (69%) partly
explains the former finding, and this is underscored by
the fact that American black patients whose disease
onset occurred befdre age 20 fared worse than Caucasians in the same group (data not shown). In the
late-onset SLE group, the probability of survival was
equivalent in Caucasians and American bIacks, and
the worse prognosis can be partly explained by the
higher frequency of atherosclerotic cardiovascular disease and neoplasia (37). This is especially important to
note, because other features of SLE that are classi-
45
cally associated with worse prognosis (e.g., renal
disease) were much less frequent in this group. In all
clinical and demographic groups considered, infection
emerged as a leading cause of death; this finding has
also been seen in other recent studies (3840).
The major impact of thrombocytopenia, another independent factor in SLE prognosis, occurred
during the first 1-2 years after disease onset. However, bleeding (and complications thereof) was an
uncommon cause of death in this group, in whom
infection emerged as the most immediate factor contributing to death. We could not demonstrate significant differences in either the prednisone dosage at the
last hospital admission or the use of immunosuppressive therapy in those with or without thrombocytopenia that might account for this; thus, thrombocytopenia may, instead, be a marker of more aggressive and
severe disease, as has been suggested by other investigators (41). Thrornbocytopenia was present with
equal frequency in those with and those without renal
involvement, although it was slightly more frequent in
those with neuropsychiatric disease, as has been
shown in other studies (10). Two other groups of
investigators (42,43) studying fairly small numbers of
patients with thrombocytopenia (n = 24 and n = 21,
respectively), could not confirm an association between this complication and more aggressive disease.
Whether the differences between these 2 studies and
our findings are due to the differences in numbers of
patients evaluated, to treatment, or to other factors
cannot be determined.
In clinical studies like the one presented here,
there are a number of sources of possible bias (44).A
study such as this might miss patients with rapidly
fatal or short disease courses, as well as those with
mild cases. We limited study inclusion to those who
met 4 or more ARA criteria for SLE, and those with
milder disease may therefore have been excluded.
Furthermore, we did not obtain followup status on
most of the 40 patients who did not meet our study
inclusion criteria (4 or more SLE classification criteria). However, the numbers of American blacks and
Caucasians who did not meet the disease criteria were
nearly the same and were fairly low (19 American
blacks and 21 Caucasians), so any small impact this
source of bias might have would be approximately the
same in both races. SLE is usually not a rapidly fatal
disease, and in this study spanning 10 years, only 3
patients died within 1 month of disease onset. Thus, it
is unlikely that a significant number of cases were
missed in this way.
46
To avoid bias toward patients with more severe
disease, we examined rheumatology outpatient records over most of this period and outpatient nephrology division records after 1982, as well as the inpatient
records. While the different methods of identifying
SLE outpatients (see Patients and Methods) might
raise concerns about confounding effects of different
sources of care and different methods of identification,
nearly all of the SLE patients in this study, regardless
of which UAB affiliated hospitals they attended, were
either directly cared for or were seen in consultation
by UAB faculty rheumatologists or nephrologists.
Only a fairly small proportion of cases were seen only
once or twice, receiving most of their medical care
elsewhere. We therefore believe that significant differences in treatment were not a major factor here. With
the different methods we used to identify outpatients,
we attempted to include all outpatients with SLE seen
at UAB over the study period. For example, a serum
bank was kept of all SLE outpatients seen at UAB by
the rheumatology division between 1975 and 1979. The
diagnoses were reviewed thereafter, and only those
meeting 4 or more classification criteria for SLE were
included in this study. We were able to ascertain fall
1985 status in all but 7% of our SLE patients, and we
believe that no significant nonresponse bias has affected our results.
A number of clinical features that were demonstrated by univariate analysis as significantly impacting
prognosis, such as renal disease and neuropsychiatric
disease and, at diagnosis, proteinuria, hypertension,
and anemia, were not significant by multivariate analysis. These variables could have been dependent on
other factors, such as black race and age at onset of
SLE. Alternatively, since the P values found for these
variables did not maintain significance when corrected
for the number of statistical tests performed, we
cannot rule out a multiple-testing bias. Some of those
parameters that were significant both on univariate
and multivariate analysis, such as black race, might
also lose significance when corrected for the number
of tests performed. However, since these were significant on both univariate and multivariate analyses,
both at diagnosis as well as at any time in the disease
course, we believe that correction for the number of
tests performed is not necessary in this setting. Last,
our calculations measured the impact of the presence
or absence of hypocomplementemia (at diagnosis) on
survival. We did not examine the relative impact of
severe versus mild depressions of total hemolytic
REVEILLE ET AL
complement, and we cannot exclude a relative difference between these two in the impact on prognosis.
Thus, our analyses of the impact of such demographic factors as race and age at onset on the clinical
presentation and prognosis in SLE showed that American black race and increasing age at SLE onset were
independent risk factors for a worse prognosis.
Thrombocytopenia, whether present at SLE diagnosis
or at any other time in the disease course, emerged as
the sole clinical parameter that independently impacted survival probability. Future studies should thus
be aimed at defining features not examined here that
contribute to a younger mean age at SLE onset in
American blacks (such as genetic or hormonal factors), as well as other confounding variables (such as
compliance and educational status) that might affect
disease outcome, so that public health measures (such
as patient education about SLE) could be taken to
improve the probability of survival in this ethnic
group. Furthermore, identification of the cause(s) of
thrombocytopenia and its specific contribution to mortality in SLE would allow a more prudent approach to
therapy in all patient groups, ultimately allowing for a
better prognosis in patients with SLE.
ACKNOWLEDGMENTS
This study would have been impossible without the
advice and support of Dr. Gene V. Ball, who participated in
the care of most of the patients. Appreciation is also extended to Dr. Marc C. Hochberg for his advice in the study
design, and to Dorothy J. Bevins for assistance in the
preparation of the manuscript.
REFERENCES
1 . Merrell M, Shulman LE: Determination of prognosis in
chronic disease, illustrated by systemic lupus erythematosus. J Chronic Dis 1:12-32, 1955
2. Rupe CE, Nickel SN: New clinical concepts of systemic
lupus erythematosus. JAMA 171:1055-1061, 1959
3. McCombs RP, Patterson JF: Factors influencing the
course and prognosis of systemic lupus erythematosus.
N Engl J Med 260:1195-1204, 1959
4. Kellum RE, Haserick JR: Systemic lupus erythematosus: a statistical evaluation of mortality based on a
consecutive series of 299 patients. Arch Intern Med
113192-99, 1964
5. Siege1 M, Gwon N , Lee SL, Rivero I, Wong W: Survivorship in systemic lupus erythematosus: relationships
to race and pregnancy. Arthritis Rheum 12:117-125,
1969
6. Estes D, Christian C: The natural history of systemic
FACTORS AFFECTING SLE PROGNOSIS
7.
8.
9.
10.
11.
12.
13.
14.
IS.
16.
17.
18.
19.
lupus erythematosus by prospective analysis. Medicine
(Baltimore) 50:85-95, 1971
Dubois EL, Wierzchowiecki M, Cox MB, Weiner JM:
Duration and death in systemic lupus erythematosus: an
analysis of 249 cases. JAMA 227:1399-1402, 1974
Fries JF, Weyls S, Holman HR: Estimating prognosis in
systemic lupus erythematosus. Am J Med 57561-565,
1974
Fessel WJ: Systemic lupus erythematosus in the community: incidence, prevalence, outcome and first symptoms: the high prevalence in black women. Arch Intern
Med 134:1027-1035, 1974
Feinglass EJ, Arnett FC, Dorsch CA, Zizic TM, Stevens
MB: Neuropsychiatric manifestations of systemic lupus
erythematosus: diagnosis, clinical spectrum, and relationship to other features of the disease. Medicine
(Baltimore) 55:323--339, 1976
Lee P, Urowitz MB, Bookman AAM, Koehler BE,
Smythe HA, Gordon DA, Ogryzlo MA: Systemic lupus
erythematosus: a review of 110 cases with respect to
nephritis, the nervous system, infection, aseptic necrosis and prognosis. Q J Med 46:l-32, 1977
Grigor R, Edmonds J , Lewkonia R, Bresnihan B,
Hughes GRV: Systemic lupus erythematosus: a prospective analysis. Ann Rheum Dis 37:121-128, 1978
Wallace DJ, Podell T, Weiner J, Klinenberg JR, Forouzesh S,, Dubois EL: Systemic lupus erythematosussurvival patterns: experience with 609 patients. JAMA
246:93&938, 1981
Hochberg MC, Dorsch CA, Feinglass EJ, Stevens MB:
Survivorship in systemic lupus erythematosus: effect of
antibody to extractable nuclear antigen. Arthritis Rheum
24:5&59, 1981
Ginzler EM, Diamond HS, Weiner M , Schlesinger M,
Fries JF, Wasner C, Medsger TA Jr, Ziegler G. Klippel
JH, Hadler NM, Albert DA, Hess EV. Spencer-Green
G, Grayzel A, Worth D, Hahn BH. Barnett EV: A
multicenter study of outcome in systemic lupus erythematodus. I: Entry variables as predictors of prognosis.
Arthritis Rheum 25:601411, 1982
Studenski S. Allen NB. Caldwell DS. Kice JR, Polisson
RP: Survival in systemic lupus erythematosus: a multivariate analysis of demographic factors. Arthritis
Rheum 30:132&1332. 1987
BallokSP, Khan MA, Kushner I: Clinical features of
systemic lupus erythematosus: differences related to
race and age of onset. Arthritis Rheum 25:55-60, 1982
Hochberg MC, Boyd RE, Ahearn JM, Arnett FC, Bias
WB, Provost TT, Stevens MB: Systemic lupus erythematosus: a review of clinico-laboratory features and
immunogenetic markers in I50 patients with emphasis
on demographic subsets. Medicine (Baltimore) 64:
285-295. 1985
Masi AT, Kaslow RA: Sex effects in systemic lupus
47
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
erythematosus: a clue to pathogenesis. Arthritis Rheum
21:480-484, 1978
Baker SB, Rovira JR, Campion EW, Mills JA: Late
onset systemic lupus erythematosus. Am J Med 66:
727-732, 1979
Wilson HA, Hamilton ME, Spyker DA, Brunner CM,
O’Brien WM, Davis JS IV, Winfield JB: Age influences
the clinical and serologic expression of systemic lupus
erythematosus. Arthritis Rheum 24: 1230-1235, 1981
Foad BSI, Sheon RP, Kirsner AB: Systemic lupus
erythematosus in the elderly. Arch Intern Med 130:
743-746, 1972
Dimant J, Ginzler EM, Schlesinger M, Diamond HS:
Systemic lupus erythematosus in the older age group:
computer analysis. J Am Geriatr SOC27:5&51, 1979
Urowitz MB, Stevens MB, Shulrnan LE: The influence
of age on clinical pattern of systemic lupus erythematosus. Arthritis Rheum 10:319-320, 1967
Catoggio LF, Skinner RP, Smith G, Maddison PS:
Systemic lupus erythematosus in the elderly: clinical
and serological characteristics. J Rheumatol 11: 175-181,
1984
Abeles M, Urman JD, Weinstein A, Lowenstein M,
Rothfield NF: Systemic lupus erythematosus in the
younger patient: survival studies. J Rheumatol 751.5522, 1980
Hochberg MC: The incidence of systemic lupus erythematosus in Baltimore, Maryland, 1970-1977. Arthritis
Rheum 28:80-86, 1985
Tan EM, Cohen AS, Fries JF, Masi AT, McShane DJ,
Rothfield NF, Schaller JG, Talal N, Winchester RJ: The
1982 revised criteria for the classification of systemic
lupus erythematosus. Arthritis Rheum 25: 1271-1277,
1982
Pincus T, Schur PH, Rose JA, Decker JL, Talal N:
Measurements of serum DNA-binding activity in SLE.
N Engl J Med 281:701-705, 1969
Aarden LA, de Groot ER, Feltkamp TEW: Immunology
of DNA-111: Crithidia luciliae, a simple substrate for the
determination of anti-dsDNA with the immunofluorescent technique. Ann NY Acad Sci 254505-515, 1975
Greenwood M: Epidemiology, Historical and Experimental. Baltimore, Johns Hopkins Press, 1932
Cox DR: Regression models and life tables. J R Stat SOC
[B] 34:187-220, 1972
US Department of Health and Human Services: Mortality, Part I, Vital Statistics of the US, 1983. Vol. 2.
National Center for Health Statistics, Hyattsville, MD,
I987
Hollingshead AB, Redlich FC: Social Class and Mental
Illness. New York, Wiley Press, 1958
Esdaile JM, Sampalis JS, Lacaille D, Danoff D: The
relationship of socioeconomic status to subsequent
health status in systemic lupus erythematosus. Arthritis
Rheum 31:423427, 1988
REVEILLE ET AL
48
36. Hochberg MC, Sutton JD: Physical disability and psychosocial dysfunction in SLE. J Rheumatol 15:959-964,
1988
37. Rubin LA, Urowitz MB, Gladman DD: Mortality in
systemic lupus erythematosus: the bimodal pattern revisited. Q J Med 216:87-98, 1985
38. Rosner S, Ginzler EM, Diamond HS, Weiner M, Schlesinger M, Fries JF, Wasner C, Medsger TA Jr, Ziegler G,
Klippel JH, Hadler NM, Albert DA, Hess EV, SpencerGreen G, Grayzel A, Worth D, Hahn BH, Barnett EV:
A multicenter study of outcome in systemic lupus
erythematosus. 11. Causes of death. Arthritis Rheum
25:612-617, 1982
39. Nived 0, Sturfelt G, Wollheim F: Systemic lupus
erythematosus and infection: a controlled and prospec-
40.
41.
42.
43.
44.
tive study including an epidemiological group. Q J Med
2181271-287, 1985
H a m s EN, Williams E, Shah DJ, de Ceulaer K: Mortality of Jamaican patients with systemic lupus erythematosus. Br J Rheumatol 28: 113-1 17, 1989
Fries JF, Holman HR: Systemic Lupus Erythematosus:
A Clinical Analysis. Philadelphia, WB Saunders, 1975
Alger M, Alarc6n-Segovia D, Rivero SJ: Hemolytic
anemia and thrombocytopenic purpura: two related subsets of systemic lupus erythematosus. J Rheumatol
4:351-357, 1977
Miller MH, Urowitz MB, Gladman DD: The significance
of thrombocytopenia in systemic lupus erythematosus.
Arthritis Rheum 26: 1181-1 186, 1983
Sackett DL: Bias in analytic research. J Chronic Dis
3251-63, 1979