вход по аккаунту


Mortality outcomes in pediatric rheumatology in the US.

код для вставкиСкачать
Vol. 62, No. 2, February 2010, pp 599–608
DOI 10.1002/art.27218
© 2010, American College of Rheumatology
Mortality Outcomes in Pediatric Rheumatology in the US
Philip J. Hashkes,1 Bridget M. Wright,2 Michael S. Lauer,3 Sarah E. Worley,1 Anne S. Tang,1
Philip A. Roettcher,4 and Suzanne L. Bowyer5
23 (21%), and were unknown in 12 patients (11%).
Rheumatic diagnoses, age at diagnosis, sex, and early
use of systemic steroids and methotrexate were significantly associated with the risk of death.
Conclusion. Our findings indicate that the overall
mortality rate for pediatric rheumatic diseases was not
increased. Even for the diseases and conditions associated with increased mortality, mortality rates were
significantly lower than those reported in previous
Objective. To describe mortality rates, causes of
death, and potential mortality risk factors in pediatric
rheumatic diseases in the US.
Methods. We used the Indianapolis Pediatric
Rheumatology Disease Registry, which includes 49,023
patients from 62 centers who were newly diagnosed
between 1992 and 2001. Identifiers were matched with
the Social Security Death Index censored for March
2005. Deaths were confirmed by death certificates, referring physicians, and medical records. Causes of
death were derived by chart review or from the death
certificate. Standardized mortality ratios (SMRs) and
95% confidence intervals (95% CIs) were determined.
Results. After excluding patients with malignancy, 110 deaths among 48,885 patients (0.23%) were
confirmed. Patients had been followed up for a mean ⴞ
SD of 7.9 ⴞ 2.7 years. The SMR of the entire cohort was
significantly decreased (0.65 [95% CI 0.53–0.78]), with
differences in patients followed up for >9 years. The
SMR was significantly greater for systemic lupus erythematosus (3.06 [95% CI 1.78–4.90]) and dermatomyositis (2.64 [95% CI 0.86–6.17]) but not for systemic
juvenile rheumatoid arthritis (1.8 [95% CI 0.66–3.92]).
The SMR was significantly decreased in pain syndromes
(0.41 [95% CI 0.21–0.72]). Causes of death were related
to the rheumatic diagnosis (including complications) in
39 patients (35%), treatment complications in 11 (10%),
non-natural causes in 25 (23%), background disease in
The practice of pediatric rheumatology includes
more than 170 conditions, both inflammatory and noninflammatory (1). Approximately 3 in 1,000 children
have a rheumatic condition (1). While we tend to study
pediatric rheumatology outcomes in terms of remission
versus active disease, organ and radiologic damage,
function, and quality of life, there is a small but significant increase in mortality rates among these patients. An
increased rate of mortality has been found in juvenile
rheumatoid arthritis (JRA) (2–12), childhood systemic
lupus erythematosus (SLE) (13–21), dermatomyositis
(DM) (22–25), various vasculitides (26–30), and systemic sclerosis (SSc) (31–33).
However, most of those studies were of relatively
small cohorts, reported mortality outcomes only on
specific diseases, had a followup time of ⬍10 years, and
were conducted prior to the 1990s, when new therapies
were developed. Even larger studies were flawed; most
were based on physician surveys and questionnaires with
no strategies to verify response accuracy (4,8). The
diagnoses in studies of national cohorts were usually not
assigned by pediatric rheumatologists (9,11). There are
no data on the mortality rate from many rare rheumatic
inflammatory diseases (primary vasculitis and autoinflammatory diseases) and noninflammatory conditions,
including pain syndromes (such as fibromyalgia). The
causes of death were usually not adequately verified, and
no systematic attempt was made to look for potential
Supported by the Northeast Ohio Chapter of the Arthritis
Philip J. Hashkes, MD, MSc, Sarah E. Worley, MS, Anne S.
Tang, MS: Cleveland Clinic, Cleveland, Ohio; 2Bridget M. Wright,
MD: The Children’s Hospital at The Medical Center of Central
Georgia, Macon; 3Michael S. Lauer, MD: National Heart, Lung and
Blood Institute, Bethesda, Maryland; 4Philip A. Roettcher, MSOR,
MBA: Indiana University, Indianapolis; 5Suzanne L. Bowyer, MD:
James W. Riley Hospital, Indianapolis, Indiana.
Address correspondence and reprint requests to Philip J.
Hashkes, MD, MSc, Consultant, Department of Rheumatic Diseases
A50, Cleveland Clinic Foundation, 9500 Euclid Avenue, Cleveland,
OH 44195. E-mail:
Submitted for publication March 25, 2009; accepted in revised
form October 16, 2009.
risk factors or predictors of mortality early in the disease
Therefore, we performed a systematic study of
mortality outcomes for all pediatric rheumatic conditions in the US based on the world’s largest pediatric
rheumatology registry. Our specific aims were to estimate the mortality rate of patients with pediatric rheumatic conditions, to describe the causes of death, and to
search for possible mortality risk factors early in the
disease course.
Indianapolis Pediatric Rheumatology Disease Registry (PRDR) data. We used the PRDR, which prospectively
collected data on new patients from 62 pediatric rheumatology
centers in the US between February 1992 and January 2001.
Data were collected on 49,023 patients at their first clinic visit,
with no followup data (34). We excluded 138 patients with a
diagnosis of malignancy, and included the remaining 48,885
patients in the analysis.
Identification of deceased patients. Patients in the
PRDR were identified only by their initials and birth date. The
full identity was known only to the referring pediatric
rheumatologist/center. Due to 2003 Health Insurance Portability and Accountability Act (HIPAA) rules prohibiting transferring data on living patients for research purposes without
specific consent (not obtained for the PRDR, which was
established in 1992), we performed the search without identifying living patients.
We matched the database list of initials and birth dates
to a list of all deaths in the US Social Security Death Index
(SSDI) corresponding to the range of birth dates in the PRDR
database. The SSDI contains the records of deceased persons
who possessed Social Security numbers and whose death had
been reported to the Social Security Administration. Figure 1
shows a flow chart of the process of determining the number of
deceased patients. The death certificates obtained from the
National Death Index of the National Center for Health
Statistics were reviewed independently by 3 physicians (PJH,
BMW, and SLB), who were blinded with regard to the original
diagnosis in the database, using standard rules for association
to a diagnosis of a rheumatic disease. These rules were based
on the cause of death and the proximity of the state where the
death occurred to the center where the patient was seen. We
identified deaths that were definitely, probably, and possibly
associated with rheumatic conditions. We then reviewed the
PRDR entry for these deceased patients to match the sex and
diagnosis to the death certificate and contacted referring
physicians for confirmation. In order to detect deceased patients who may have been missed in our search, we performed
a sensitivity analysis by asking all referring center physicians
whether they were aware of any deaths among former patients.
We then checked whether they had been reported to the
PRDR. Only 5 deceased patients were identified by this
method, and 4 of these were from 1 center. The medical charts
of 87 deceased patients (79%) were reviewed to confirm the
Figure 1. Flow chart showing the process of detection of deceased
patients from the Indianapolis Pediatric Rheumatology Disease Registry (PRDR). Definite, probable, and possible refer to deaths that
were definitely, probably, and possibly associated with rheumatic
initial diagnosis, cause of death, and demographic, clinical, and
treatment factors related to these patients.
Identification of causes of death. The cause of death
was obtained from the patient’s chart and/or autopsy, when
available. For other patients we used the cause stated in the
death certificate. The cause of death was independently classified by 4 physicians (PJH, BMW, MSL, and SLB) into causes
related to the rheumatic diagnosis (including disease complications), treatment complications, non-natural causes (e.g.,
accidents, homicide, and suicide), background disease, and
unknown/unclear causes. The certainty of the cause of death
was classified into definite (for example, autopsy-proven,
microbiology-proven, or non-natural deaths), probable (clinical circumstantial evidence), possible (lack of evidence other
than death certificate), and unknown.
Identification of potential risk factors/predictors of
mortality. The PRDR included data for almost all patients on
demographic (age, sex, and ethnicity) and diagnostic factors
(rheumatic diagnoses based on International Classification of
Diseases, Ninth Revision codes; time of disease onset; and
diagnosis). There were additional queries that changed every
1–2 years (34). We obtained data on patient zip code and
medical insurance as surrogates for socioeconomic data.
(Other surrogates were missing from almost all charts.)
Statistical analysis. Expected survival for the entire
cohort, for specific disease categories, specific diseases, and
inflammatory versus noninflammatory conditions was computed from the age- and sex-matched US population and
compared with observed survival using the 1-sample log rank
test (35,36); for nearly 10% of the patients information on
ethnicity was missing, so this factor was not used in this
computation. The standardized mortality ratio (SMR) comparing observed to expected survival was computed, with 95%
confidence intervals (95% CIs) calculated from the Poisson CI
on the number of deaths.
The association between survival times and potential
risk factors was assessed using a separate Cox proportional
hazards model for each risk factor, with time to death as the
dependent variable, censored at March 1, 2005 (6 months prior
to the SSDI search date of September 1, 2005), since the SSDI
is sensitive in detecting deaths up to 6 months prior to this date
(37,38). Time 0 was defined as the date of clinic visit rather
than the date of onset of disease, since the date of onset is
often a subjective judgment and was not available for many
To assess predictors of survival multivariably while
adjusting for and assessing the potential correlation between
the outcomes of subjects treated at the same center, we used
the robust sandwich estimate for the covariance matrix (39).
Only variables for data that were missing on ⬍5% of observations and, in the case of categorical variables, had ⱖ5% of
observations in each level of the variable were used in the
modeling process. The appropriateness of the proportional
hazards assumption was assessed graphically using log–log
survival plots and by entering risk factor–by-time interactions
into the model, and model fit was assessed using analysis of
residuals. The final model was limited to 11 variables (10% of
110 events) to avoid overparameterization. To improve the
robustness of our model, parameter estimates to the effects of
influential observations, the final model was fit using 1,000
samples bootstrapped from the data with replacement, and the
mean, 2.5th percentile, and 97.5th percentile of the parameter
estimates from those models were used to obtain final esti-
mates and 95% CIs for the hazard ratios (HRs) of the risk
For review of the medical charts of the deceased
patients, we used descriptive statistics. We obtained median
household incomes for US census data for deceased patients’
zip codes of residence and compared the distribution of these
median incomes to 2006 household income quintiles for the
US using a chi-square goodness-of-fit test (http://www.census.
All analyses used complete cases, and all tests were
2-tailed and performed at a significance level of 0.05. SAS,
version 9.2 (SAS Institute, Cary, NC) and R 2.3.1 (R Foundation for Statistical Computing, Vienna, Austria) were used for
analysis. Oracle and SQL software were used for the SSDI and
PRDR list match.
Summary data from the PRDR. Data were collected on 48,885 patients (34). There were 56,260 rheumatic diagnoses in 48,322 patients, with 6,947 patients
having more than one diagnosis. Of the 39,221 patients
for whom information was available, 24,911 (63.5%) had
chronic inflammatory diseases. The most common inflammatory diagnoses were JRA (in 9,894 of the 24,911
patients with a diagnosis of inflammatory disease
[39.7%]), SLE (in 1,440 [5.8%]), Raynaud’s phenomenon (in 1,222 [4.9%]), Henoch-Schönlein purpura (HSP;
in 849 [3.4%]), juvenile DM (in 686 [2.8%]), and scleroderma (in 419 [1.7%]). The most common noninflammatory diagnoses included arthralgia (in 5,112 of the
14,310 patients with noninflammatory disease [35.7%]),
antinuclear antibody positivity (in 4,346 [30.4%]), hyper-
Table 1. Summary statistics from the Indianapolis Pediatric Rheumatology Disease Registry cohort*
Entire cohort
Age at visit, years
Age at death, years
Time from onset to diagnosis, months
Weight, kg
Height, cm
Time to travel to medical center, minutes†
Distance from medical center, miles†
Sex, no. (%) female/male
Ethnicity, no. (%)
African American
Native American
10.0 ⫾ 4.7 (10.3 [0–28]); 48,207
12.6 ⫾ 20.3 (5.0 [0–243.2]); 27,352
38.8 ⫾ 21.4 (35.0 [1–179]); 6,794
133.7 ⫾ 27.1 (136.0 [52.5–200]); 2,865
64.1 ⫾ 64.1 (45.0 [0–1,440]); 6,618
81.6 ⫾ 480.2 (25.0 [0–10,000]); 641
30,865 (64.2)/17,181 (35.8)
35,032 (79.5)
5,121 (11.6)
2,297 (5.2)
732 (1.7)
59 (0.1)
831 (1.9)
Deceased patients
11.6 ⫾ 5.1 (12.7 [0.2–21.6]); 110
16.6 ⫾ 6.1 (17.4 [0.9–30.4]); 110
10.6 ⫾ 14.3 (4.5 [1.0–76.1]); 56
39.2 ⫾ 19.7 (42.0 [12.0–71.0]); 9
160.3 ⫾ 11.3 (160.3 [152.3–168.3]); 2
73.3 ⫾ 92.4 (20 [20–180]); 3
40.0 ⫾ 10.0 (40 [30–50]); 3
62 (56.4)/48 (43.6)
75 (71.4)
21 (20)
7 (6.7)
2 (1.9)
0 (0)
0 (0)
* Except where indicated otherwise, values are the mean ⫾ SD (median [range]); number of patients. The earliest date of birth of patients in the
registry was February 16, 1970, and the latest date of birth was May 31, 2001 (for 48,427 patients). The earliest date of visit was January 11, 1990,
and the latest date of visit was November 2, 2001 (for 48,721 patients). The earliest date of death was March 2, 1992, and the latest date of death
was February 18, 2005 (for 110 patients). Ethnicity was unknown for 5 of the 110 deceased patients.
† Some of the patients traveled from outside the US.
Table 2. SMRs for the entire cohort and for diagnosis subgroups*
Observed deaths,
no. (%)
SMR (95% CI)
110 (0.23)
17 (1.2)
5 (0.8)
19 (0.2)
7 (0.2)
6 (0.6)
12 (0.1)
5 (0.2)
4 (0.1)
4 (1.9)
5 (0.2)
3 (0.1)
3 (1.4)
6 (0.1)
6 (0.3)
24 (0.5)
14 (0.4)
64 (0.3)
29 (0.2)
0.65 (0.53–0.78)
3.06 (1.78–4.90)
2.64 (0.86–6.17)
0.57 (0.34–0.89)
0.43 (0.17–0.88)
1.8 (0.66–3.92)
0.41 (0.21–0.72)
0.59 (0.19–1.38)
0.23 (0.06–0.58)
4.71 (1.28–12.07)
0.49 (0.16–1.14)
0.29 (0.06–0.84)
3.47 (0.72–10.15)
0.46 (0.17–0.99)
0.69 (0.25–1.51)
1.22 (0.78–1.81)
1.04 (0.57–1.74)
0.76 (0.58–0.97)
0.58 (0.39–0.83)
Entire cohort (n ⫽ 47,449)
SLE (n ⫽ 1,393)
Dermatomyositis (n ⫽ 662)
JRA (all) (n ⫽ 9,604)
Other arthritis (n ⫽ 4,614)
Systemic JRA (n ⫽ 962)
Pain syndromes (all) (n ⫽ 8,147)
Fibromyalgia (n ⫽ 2,297)
Arthralgia (n ⫽ 5,324)
Primary vasculitis (n ⫽ 206)†
Orthopedic/mechanical disorders (n ⫽ 2,401)
Infection (n ⫽ 2,857)
Genetic/chromosomal/metabolic/bone dysplasia (n ⫽ 208)
Laboratory abnormalities (n ⫽ 4,466)
Other rheumatic diagnosis (n ⫽ 2,286)
Other nonrheumatic diagnosis (n ⫽ 5,316)
Diagnosis not clear/unspecified (n ⫽ 3,829)
Inflammatory disease (n ⫽ 24,187)
Noninflammatory disease (n ⫽ 13,920)
* Only patients with complete data on visit date, date of birth, and sex were included, so the numbers of patients do not equal the overall number
of patients in the Indianapolis Pediatric Rheumatology Disease Registry. This table includes data primarily on disease categories. For specific
diseases/disease subtypes we included those we thought would be of particular interest. For other diseases/subtypes the standardized mortality
ratio (SMR) was not statistically significant. The SMR is defined as the number of observed deaths divided by the number of expected deaths. 95%
CI ⫽ 95% confidence interval; SLE ⫽ systemic lupus erythematosus; JRA ⫽ juvenile rheumatoid arthritis.
† Except for Kawasaki disease and Henoch-Schönlein purpura.
Figure 2. Kaplan-Meier survival curves for the entire cohort (n ⫽ 47,449) (A), for patients with systemic lupus erythematosus (n ⫽ 1,393) (B),
for patients with systemic juvenile rheumatoid arthritis (n ⫽ 962) (C), and for patients with primary vasculitis other than Kawasaki disease or
Henoch-Schönlein purpura (n ⫽ 206) (D). These plots include only patients for whom complete data on date of visit, date of birth, and sex were
available, so the numbers of patients do not equal the overall numbers in the Indianapolis Pediatric Rheumatology Disease Registry.
mobility syndrome (in 2,416 [16.9%]), and fibromyalgia
(in 2,234 [15.6%]).
The SSDI search was performed a mean of 7.9 ⫾
2.7 years after the patients were registered (median 7.8
Table 3. Causes of death by diagnostic category*
Cause of death
Related to diagnosis
assigned by a
(or other) disease
JRA and other chronic arthritis (n ⫽ 20)
Connective tissue diseases (n ⫽ 26)
Vasculitis (n ⫽ 5)
Infection (n ⫽ 3)
Orthopedic/mechanical/bone dysplasia (n ⫽ 2)
Pain syndromes (n ⫽ 17)
Laboratory abnormalities (n ⫽ 2)
Other rheumatic diagnosis (n ⫽ 2)
Other nonrheumatic diagnosis (n ⫽ 14)
7 (35)
15 (58)
4 (80)
0 (0)
1 (50)
0 (0)
0 (0)
0 (0)
10 (71)
4 (20)
5 (19)
1 (20)
1 (33)
0 (0)
0 (0)
0 (0)
0 (0)
0 (0)
5 (25)
2 (8)
0 (0)
2 (67)
1 (50)
10 (59)
1 (50)
1 (50)
0 (0)
4 (20)
4 (15)
0 (0)
0 (0)
0 (0)
7 (41)
1 (50)
1 (50)
4 (29)
* Values are the number (%) of patients. Numbers of patients differ slightly from those shown in other tables due to differences between the
Indianapolis Pediatric Rheumatology Disease Registry diagnosis (which includes ⬎1 diagnosis for some deceased patients) and the diagnosis
obtained from the chart review. Only the primary diagnosis was used in this analysis. JRA ⫽ juvenile rheumatoid arthritis.
[range 0–14.6 years]), with no significant differences
between the various diagnostic categories. Other summary data are detailed in Table 1.
Mortality rates. We identified 110 deceased subjects (0.23% [95% CI 0.19–0.27]). The mortality rate of
the PRDR cohort was significantly lower than the
expected mortality rate from the age- and sex-adjusted
US population (Table 2 and Figure 2A). However,
significant differences in observed and expected mortality rates were seen only among the 18,111 patients who
were followed up for at least 9 years.
For several diagnostic categories and specific
diseases, we found a significant change in the population
SMR (Table 2 and Figures 2B–D). Among disease
categories significant increases in the population SMR
and in mortality rates were found for connective tissue
diseases and primary vasculitis (except for Kawasaki
disease and HSP) when compared with the rest of the
PRDR cohort. The SMR and the mortality rate were
significantly decreased for pain syndromes and arthralgia compared with the rest of the PRDR cohort. Among
specific diseases the SMR was significantly increased for
SLE and DM. For other specific diseases, including all
subtypes of JRA, Kawasaki disease, and HSP, the SMR
and the mortality rate did not significantly differ from
that of the rest of the PRDR cohort. Sixty-four (58%) of
the deaths were in patients with a primary diagnosis of
inflammatory disease, and 45 (41%) were in patients
with a primary diagnosis of noninflammatory disease.
(The primary diagnosis was unknown in 1 patient.) The
SMRs of both the noninflammatory and inflammatory
groups were significantly lower than the general population (but the difference was more significant for the
noninflammatory group).
The 5-year survival rate was 99.9% (95% CI
99.8–99.9) for the entire cohort, 99.5% (95% CI 99.1–
99.9) for SLE, 99.6% (95% CI 99.4–99.9) for all connective tissue diseases, 99.8% (95% CI 99.5–100) for systemic JRA, and 99.0% (95% CI 97.7–100) for primary
vasculitis other than Kawasaki disease and HSP (Figures
2A–D). The 10-year survival rate was 99.7% (95% CI
99.6–99.8) for the entire cohort, 98.2% (95% CI 97.2–
99.2) for SLE, 98.8% (95% CI 98.2–99.3) for all connective tissue diseases, 99.1% (95% CI 98.3–99.8) for
systemic JRA, and 96.7% (95% CI 93.2–100) for primary vasculitis other than Kawasaki disease and HSP.
The mean time to death for the entire cohort was 5.0 ⫾
3.4 years (median 4.9 [range 0–12.2 years]). No significant differences were seen between those with inflammatory conditions and those with noninflammatory conditions.
Causes of death. The causes of death were related to the diagnosis for which the deceased was seen by
a rheumatologist (including complications) in 39 patients (35%), treatment complications in 11 (10%),
non-natural causes in 25 (23%), background disease in
23 (21%), and were unknown/unclear in 12 (11%). The
cause of death was ascertained from the chart in only 29
patients (26%); among these 11 (38%) were confirmed
by an autopsy. The cause of death for the other 81
patients (74%) was obtained from the death certificate,
except for 3 patients, for whom no cause of death was
available. The cause of death was definitive for 47
patients (43%), probable for 31 patients (28%), possible
for 23 patients (21%), and unknown or unclear for 9
patients (8%).
There was a significant correlation between the
rheumatic diagnosis and the cause of death (Table 3).
Among the 64 deceased patients in whom inflammatory
disease was diagnosed, 28 (44%) died of a cause related
to the diagnosis for which they were seen by a rheumatologist, while only 11 (24%) of the 45 patients in whom
a noninflammatory disease was diagnosed died of a
cause directly related to the diagnosis. In contrast, 13
(29%) and 14 (31%), respectively, of the patients in
whom a noninflammatory disease was diagnosed died of
a non-natural cause or of a background (or other)
disease, as opposed to 11 (17%) and 9 (14%), respectively, among patients in whom an inflammatory disease
was diagnosed (P ⬍ 0.001). This was particularly evident
for patients diagnosed with a pain syndrome, among
whom 59% died of non-natural causes and 41% of
background (or other) disease.
Among patients with SLE, 6 died of renal disease, 2 of pancreatitis, 1 of pulmonary hemorrhage, and
1 of intracranial hemorrhage. Four patients with SLE
died of infection; of these, 2 patients had had bone
marrow transplantation. Two patients with systemic
JRA died of macrophage activation syndrome, 2 of heart
disease, 1 of infection, and 1 of a secondary malignancy.
Two patients with DM died of heart disease, 1 of a
myocardial infarction, and 1 of myocarditis. One died of
aspiration pneumonia, 1 of gastrointestinal perforation,
and 1 of a surgical complication related to aortic coarctation (DiGeorge syndrome). Six patients with pain
syndromes committed suicide, and 2 died of homicide.
Among the patients with vasculitis who died, most
causes of death were related to their disease, but causes
were well described for only 2 patients. One patient with
polyarteritis nodosa died of a myocardial infarction, and
a patient with HSP died of gastrointestinal perforation
and renal failure.
Predictors of mortality. When comparing characteristics of deceased patients with survivors in the
PRDR, only the age at visit and distance from the center
were significant factors. An older age at the time of visit
and less distance from the pediatric rheumatology center
were associated with increased mortality (Table 1). The
other factors that we examined, including sex, ethnicity,
time from onset of disease to diagnosis, weight, height,
initial medication use, and time required to get to the
treating center were not significantly associated with
mortality risk.
In univariable survival analysis, several diagnostic
categories and diseases were significant predictors of a
higher risk of mortality. These included all connective
tissue diseases (HR 4.5 [95% CI 2.9–6.9]), SLE (HR 6.0
[95% CI 3.6–10.1]), DM (HR 3.3 [95% CI 1.3–8.0]),
primary vasculitis, except for Kawasaki disease and HSP
Table 4. Significant HRs for diagnoses and other predictors of
mortality versus survival in a multivariable survival model*
HR (95% CI)
Older age at visit (⬎14.5 years)
Pain syndromes
Connective tissue disease
Juvenile rheumatoid arthritis
Other arthritis
Other nonrheumatic diagnosis
Unspecified diagnosis
Male sex
2.3 (1.6–3.3)
0.43 (0.09–0.97)
0.83 (0.37–1.5)
4.8 (2.8–8.4)
0.55 (0.00–1.5)
1.2 (0.62–2.2)
0.83 (0.25–1.8)
2.5 (1.4–4.0)
2.0 (0.92–3.9)
1.7 (1.1–2.5)
* HRs ⫽ hazard ratios; 95% CI ⫽ 95% confidence interval.
(HR 9.2 [95% CI 3.4–25.0]), systemic JRA (HR 2.5
[95% CI 1.1–5.7]), and genetic/chromosomal/metabolic
diseases (HR 6.2 [95% CI 2.0–19.5]). Diagnoses of
arthralgia (HR 0.3 [95% CI 0.1–0.9]) were significantly
predictive of better survival; pain syndromes were marginally predictive (HR 0.6 [95% CI 0.3–1.1]). Other
significant predictors of mortality included an older age
at the time of the first visit to a rheumatologist (HR 1.1
[95% CI 1.0–1.1] per year of age) and the use of systemic
steroids (HR 5.6 [95% CI 1.1–28.9]) or methotrexate
(HR 15.1 [95% CI 2.9–77.7]) at the initial visit. Other
disease categories and specific diseases, sex, ethnicity,
height, weight, year of visit, distance and time from
rheumatology center, time from diagnosis, erythrocyte
sedimentation rate, and presence of antinuclear antibody were not associated with increased mortality risk.
The number of patients was too small for calculation of
HRs for other factors in the PRDR.
In the multivariable model only connective tissue
diseases, other nonrheumatic diagnosis, male sex (even
though not significant in the univariable analysis), and
older age at visit were significantly associated with
mortality, while arthralgia was a negative predictor of
mortality (Table 4).
Review of the charts of deceased patients. We
reviewed charts for 87 patients (79%) (5 only partially).
Seventeen charts (15%) were not located or had been
destroyed. For 3 patients there were no pediatric rheumatologists currently in that locality, and for 3 patients
there was lack of cooperation or institutional review
board refusal to allow chart review. The deceased patients had resided in 30 states, with the mode from Ohio
(17 patients [15%]). These numbers correlated with the
number of patients submitted to the PRDR from each
state (data not shown). Thirty (27%) of the deceased
patients were between ages 17 and 21 years (“transition”
years); of these 15 (50%) died of disease complications
or infections, and 10 (33%) died of non-natural causes.
Among the 71 deceased patients for whom data
were available, 58 (82%) lived in urban localities, and 13
(18%) lived in rural localities. Of the 76 deceased
patients for whom zip code of residence was known
(69%), 2 (3%) lived in zip codes with median household
income in the lowest socioeconomic quintile, 36 (47%)
lived in zip codes with median income in the second
quintile, 32 (42%) lived in zip codes with median income
in the middle quintile, 6 (8%) lived in zip codes with
median income in the fourth quintile, and none lived in
zip codes with median income in the highest quintile.
These patients were significantly more likely to reside in
zip codes with middle-quintile median household incomes than the general population (P ⬍ 0.001).
We obtained insurance information on 79 (72%)
of the deceased patients; 53 (67%) had commercial
insurance/pay-for-service and 26 (33%) had Medicaid
(or lacked insurance). The proportion of deceased patients who had been on Medicaid was similar to that
reported for the general population for the study period
( Data on clinical and laboratory parameters and treatments are available from the corresponding author upon request.
This is the largest systematic mortality outcome
study in pediatric rheumatology published to date. Overall, we did not detect an increase in mortality compared
with the general population, and even for those diseases
and conditions associated with increased mortality the
rates were significantly lower than those reported in
previous studies, especially for systemic JRA, SLE, DM,
and vasculitis.
Previous studies of JRA showed a decreasing
mortality rate between the early 1970s and early 1990s,
but still demonstrated a significantly increased SMR
(7–11). In the literature on childhood SLE, 5-year
survival rates range from 83% to 95%, and 10-year
survival rates range from 76% to 95% (13–21). The most
recent study of juvenile DM demonstrated a mortality
rate of ⬃1.5–2.5% over 3 to 5 years (25). Mortality rates
were very low for HSP and Kawasaki disease (28,29) but
very high for polyarteritis nodosa (30). For SSc the
5-year survival rate is ⬃90%, the 10-year survival rate is
80%, and the 20-year survival rate is 69% (31–33).
One possible cause of the increased survival in
the present study compared with previous studies may
be the improved treatment that was introduced in the
1990s. Alternative reasons may be the relatively short
mean followup time in the cohort examined in the
present study and the underestimation of mortality due
to methodology limitations (discussed below).
Patients with arthralgia in the present study had a
decreased mortality rate. The reason for this is not clear,
but many of these patients see numerous physicians
(including gastroenterologists, neurologists, psychologists, and psychiatrists), and perhaps increased medical
vigilance was related to the decreased mortality rate
(40). Adult studies have shown an increased mortality in
pain syndromes (41).
As expected, most of the patients with inflammatory disease died of their disease or disease complications. With longer followup this proportion may change,
given the possibility of secondary malignancies or an
increased rate of infections related to prolonged immunosuppression. Interestingly, most of the patients who
had pain syndromes died of non-natural causes. Many
children with pain syndromes have mood disorders, such
as anxiety and/or depression, that may result in an
increased incidence of death from non-natural causes,
such as suicide, homicide, and even motor vehicle accidents resulting from suicidal intent or substance abuse.
We were unable to obtain data from these patients’
charts to support this hypothesis. Unlike in earlier
studies (2,3,6,9), no patients with JRA in the present
study died of amyloidosis, late development of other
autoimmune disease, or suicide (9,10). Two patients
with systemic JRA died of macrophage activation syndrome, which is probably currently the most common
cause of death in patients with systemic JRA (12).
With regard to early factors associated with mortality, we found an increased risk of mortality among
male patients and among patients first seen at an older
age. These factors remained significant even after adjustment for diagnosis in a multivariable model. We do
not have an explanation for these findings, although
population mortality rates in children are slightly greater
in males than in females.
An interesting finding was the increased mortality in patients living closer to the medical center, although data were available for a small fraction of the
PRDR cohort (insufficient for inclusion in the predictor
model). One possible explanation may be the existence
of many tertiary hospitals (where most pediatric rheumatologists practice) in inner city neighborhoods. Alternatively, being able to travel from a great distance to a
tertiary medical center may be a surrogate indicator of
higher socioeconomic status, as was found in other
mortality studies (42). According to a recent pediatric
rheumatology workforce analysis, the mean distance
traveled by patients to a pediatric rheumatologist is 57
miles (43). In our limited data we did not find that
socioeconomic factors (median income in the zip code of
residence and proportion of patients on Medicaid)
played a major role in mortality.
This study has limitations that may have resulted
in underestimation of mortality. External validity was
shown, however, by the survival of patients in our cohort
who were eventually diagnosed with malignancies, which
were not included in the primary analysis since they are
not considered pediatric rheumatic conditions, being
within the range of modern childhood malignancy studies (44,45). Of 138 patients with malignancy, 17 died.
The 5-year survival rate was 90.1% (95% CI 85.1–95.4),
and the 10-year survival rate was 85.2% (95% CI 78.6–
Face validity was also evident from the finding
that patients with inflammatory conditions had a higher
mortality rate than those with pain syndromes. Our
ability to detect deceased patients was decreased due to
the limited identifiers of patients in the PRDR. Directly
approaching physicians who had contributed to the
PRDR for their list of deceased patients would have
been less accurate (including physician memory bias)
and may have led to compliance issues (e.g., retired
physicians, patients that had moved, locating patients on
physician master lists, and time constraints). We could
not ask physicians to identify their complete list of
patients due to HIPAA regulations that prohibit research on living patients without specific consent. A
recent publication of the Institute of Medicine pointed
out the difficulties HIPAA rules impose on research
Identifying the deceased via the SSDI also has
limitations, since patients without a Social Security
number were missed. These may include some foreignborn patients (such as illegal immigrants), Native Americans (a small number in our cohort), and patients whose
guardians or relatives had not requested death benefits
or who had changed their name/initials (due to marriage). Despite these limitations, several studies have
shown the SSDI to be 86–97.5% sensitive and 99%
specific when searched for information on deaths that
occurred up to 6 months prior to the search date
(37,38,47). In addition, a Social Security number is
obtained for nearly all children in the US in the first year
of life (since it is needed for child exemptions on tax
forms), and most families utilize Social Security death
benefits. Pediatric rheumatic diseases and mortality are
very rare in the first year of life (1). We also performed
a sensitivity analysis (as described in Patients and Methods) for patients who were not detected in the search
results and found only 5 additional deceased patients.
Since, as our study demonstrated, mortality is relatively
uncommon in the individual pediatric rheumatology
practice, it is expected that most physicians would remember their deceased patients from the last 10–12
There may be problems in generalizing our data.
Many US centers did not participate in the PRDR, and
the compliance of participating centers in submitting
patients varied considerably. However, the PRDR,
which includes ⬃49,000 patients, is the world’s largest
registry for these conditions. Since the PRDR mainly
includes patients first seen in pediatric rheumatology
clinics, some rheumatic diseases, especially those with a
hospital-based presentation (such as SLE) or those seen
frequently by other pediatric specialists (such as Kawasaki disease, rheumatic fever, and HSP) may be
underrepresented if they were not followed up in the
clinic or registered during clinic followup. Since this
study was not a population study, it is possible that cases
with mild disease that were not referred to pediatric
rheumatologists were missed. This referral bias may
offset other factors that may have resulted in an underestimation of mortality.
The mortality rate and causes of death may not
be generalized to other parts of the world. Studies have
shown higher rates of JRA-associated amyloidosis in
Europe and suicide among patients with JRA in Finland
(2,3,9). Risk factors may differ between countries. However, differences in areas such as approach to treatment
have diminished recently with improved international
information sharing. The earliest date of death in our
cohort was 1992, when the incidence of amyloidosis was
already decreasing in Europe and international cooperation was increasing. Future studies should compare our
results with results from existing registries in other
countries (48,49).
Due to the relatively short followup period, we
did not capture the full extent of mortality, especially
late deaths after early adulthood. Thus, we may have
missed the mortality risk of premature cardiovascular
diseases known to occur in SLE and RA.
A potential strength of the present study is the
accuracy of the diagnoses, since patients were diagnosed
by qualified pediatric rheumatologists. However, patients’ diagnoses may have changed since the first visit.
Indeed, in 36 (41%) of the 87 deceased patients whose
charts were reviewed in detail the diagnosis was
changed. Also, for ⬃15% of the patients in the PRDR,
there was no definitive diagnosis. This limited our ability
to use database diagnoses for analysis. Unfortunately,
we were not able to sample the charts of the surviving
patients for diagnostic accuracy, as previously described
Sources of the cause of death are often inaccurate, including death certificates, which often report
ill-defined causes of death. In only a minority of cases we
confirmed the cause from patient charts and/or autopsy
reports. However, we were able to report a high degree
of certainty of the cause of death for 71% of the
Since the information in the PRDR was limited,
we could not explore in depth for risk factors or early
predictors of mortality. For many factors data were
available for only a minority of patients. The chart
review was also limited by lack of complete data for each
patient. Other factors should be investigated in future
registries or in case–control studies.
While the results of our study are encouraging,
with the mortality rate of our entire cohort similar to
that of the age- and sex-matched US population, it is
important to follow up this cohort in the future for
mortality trends, especially later deaths seen as sequelae
in many rheumatic diseases.
Medical School, Durham, NC), Thomas J. A. Lehman, MD
(Hospital for Special Surgery, New York, NY), Calvin B.
Williams, MD, PhD (Milwaukee Children’s Hospital, Milwaukee, WI), Beth S. Gottlieb, MD, MS (Schneider Children’s
Hospital, New Hyde Park, NY), Deborah Rothman, MD, PhD
(Shriner’s Hospital, Springfield, MA), David M. Siegel, MD,
MPH (Strong Memorial Hospital, University of Rochester,
Rochester, NY), Paula W. Morris (Arkansas Children’s Hospital, Little Rock, AR), Leonard D. Stein, MD (University of
North Carolina, Chapel Hill), Donald P. Goldsmith, MD (St.
Christopher’s Hospital, Philadelphia, PA), Linda WagnerWeiner, MD (La Rabida Hospital, University of Chicago,
Chicago, IL), Richard K. Vehe, MD (Gillette Children’s
Hospital, University of Minnesota, St. Paul, MN), Kathleen M.
O’Neil, MD (Children’s Hospital of Buffalo, Buffalo, NY),
Lawrence S. Zemel, MD (Connecticut Children’s Medical
Center, Hartford, CT). We also thank Ms Christine Skibinski
who formed and helped operate the database.
All authors were involved in drafting the article or revising it
critically for important intellectual content, and all authors approved
the final version to be published. Dr. Hashkes and Ms Worley had full
access to all of the data in the study and take responsibility for the
integrity of the data and the accuracy of the data analysis.
Study conception and design. Hashkes, Wright, Lauer, Worley.
Acquisition of data. Hashkes, Wright, Roettcher, Bowyer.
Analysis and interpretation of data. Hashkes, Lauer, Worley, Tang.
We wish to thank the following investigators of the
PRDR, who assisted us in identifying the deceased and helped
us in the chart review process, including obtaining approval by
Institutional Review Boards, hosting the authors during the
chart review visit, and reviewing several of the charts: Hermine
Brunner, MD, MSc (Cincinnati Children’s Hospital Medical
Center, Cincinnati, OH), Gloria C. Higgins, MD, PhD (Children’s Hospital of Columbus, Columbus, OH), Larry B. Vogler, MD (Egleston Children’s Hospital at Emory University
[ARPOC], Atlanta, GA); Carol A. Wallace, MD (Seattle
Children’s Hospital Center, Seattle, WA), Jorge LopezBenitez, MD (Floating Children’s Hospital, Boston, MA),
Donna L. Gibbas, MD (ARPRC, Atlanta, GA), Carol B.
Lindsley, MD (University of Kansas Medical Center, Kansas
City, KS), J. Kenneth Herd, MD (East Tennessee State,
Johnson City, TN), Marisa S. Klein-Gitelman, MD, MPH
(Children’s Memorial Hospital, Chicago, IL), Terry L. Moore,
MD (St. Louis Medical Center, St. Louis, MO), Linda K.
Myers, MD (University of Tennessee, Memphis, TN), Harry L.
Gewanter, MD and Eugenio Monasterio, MD (Children’s
Hospital of Richmond, Richmond, VA), Kenneth N. Schikler,
MD (Kosair Children’s Hospital, Louisville, KY), David
Sherry, MD and Terri H. Finkel, MD, PhD (Children’s
Hospital of Philadelphia, Philadelphia, PA), Andreas A. Reiff,
MD (Children’s Hospital of Los Angeles, Los Angeles, CA),
Ilona S. Szer, MD (Children’s Hospital of San Diego, San
Diego, CA), Egla C. Rabinovich, MD, MPH (Duke University
1. Petty RE, Cassidy JT. Introduction to the study of rheumatic
diseases in children. In: Cassidy JT, Petty RE, editors. Textbook of
pediatric rheumatology. 4th ed. Philadelphia: WB Saunders; 2001.
p. 4–8.
2. Laaksonen AL. A prognostic study of juvenile rheumatoid arthritis: analysis of 544 cases. Acta Paediatr Scand 1966;Suppl 166:
3. Ansell BM, Wood PH. Prognosis in juvenile chronic polyarthritis.
Clin Rheum Dis 1976;2:397–412.
4. Baum J, Gutowska G. Death in juvenile rheumatoid arthritis.
Arthritis Rheum 1977;20 Suppl:253–5.
5. Bernstein B. Death in juvenile rheumatoid arthritis. Arthritis
Rheum 1977;20 Suppl:256–7.
6. Bywaters EG. Death in juvenile chronic polyarthritis. Arthritis
Rheum 1977;20 Suppl:256.
7. Stoeber E. Prognosis in juvenile rheumatoid arthritis: follow-up of
433 chronic rheumatoid children. Eur J Pediatr 1981;135:225–8.
8. Wallace CA, Levinson JE. Juvenile rheumatoid arthritis: outcome
and treatment for the 1990s. Rheum Dis Clin North Am 1991;17:
9. Savolainen HA, Isomaki HA. Decrease in the number of deaths
from secondary amyloidosis in patients with juvenile rheumatoid
arthritis. J Rheumatol 1993;20:1201–3.
10. French AR, Mason T, Nelson AM, O’Fallon WM, Gabriel SE.
Increased mortality in adults with a history of juvenile rheumatoid arthritis: a population-based study. Arthritis Rheum 2001;44:
11. Thomas E, Symmonds DP, Brewster DH, Black RJ, Macfarland
GJ. National study of cause-specific mortality in rheumatoid
arthritis, juvenile chronic arthritis, and other rheumatic conditions:
a 20 year followup study. J Rheumatol 2003;30:958–65.
12. Stephan JL, Kone-Paut I, Galambrun C, Mouy R, Bader-Meunier
B, Prieur AM. Reactive hemophagocytic syndrome in children
with inflammatory disorders: a retrospective study of 24 patients.
Rheumatology (Oxford) 2001;40:1285–92.
Meislin AG, Rothfield NF. Systemic lupus erythematosus in
childhood: analysis of 42 cases with comparative data on 200 adult
cases followed concurrently. Pediatrics 1968;42:37–49.
Walravens PA, Chase HP. The prognosis of childhood systemic
lupus erythematosus. Am J Dis Child 1976;130:929–33.
Kaslow RA, Masi AT. Age, sex, and race effects on mortality from
systemic lupus erythematosus in the United States. Arthritis
Rheum 1978;21:473–9.
Abeles M, Urman JD, Weinstein A, Lowenstein M, Rothfield NF.
Systemic lupus erythematosus in the younger patients: survival
studies. J Rheumatol 1980;7:515–22.
Glidden RS, Mantzouranis EC, Borel Y. Systemic lupus erythematosus in childhood: clinical manifestations and improved survival
in fifty-five patients. Clin Immunol Immunopathol 1983;29:
Lehman TJ, McCurdy DK, Bernstein BH, King KK, Hanson V.
Systemic lupus erythematosus in the first decade of life. Pediatrics
Lacks S, White P. Morbidity associated with childhood systemic
lupus erythematosus. J Rheumatol 1990;17:941–5.
Tucker LB, Menon S, Schaller JG, Isenberg DA. Adult and
childhood onset systemic lupus erythematosus: a comparison of
onset, clinical features, serology and outcome. Br J Rheumatol
Ward MM, Pyun E, Studenski S. Causes of death in systemic lupus
erythematosus: long-term followup of an inception cohort. Arthritis Rheum 1995;38:1492–99.
Bitnum S, Daeschner CW Jr, Travis LB, Dodge WF, Hopps HC.
Dermatomyositis. J Pediatr 1964;64:101–31.
Bowyer SL, Blane CE, Sullivan DB, Cassidy JT. Childhood
dermatomyositis: factors predicting functional outcome and development of dystrophic calcinosis. J Pediatr 1983;103:882–8.
Hochberg MC. Epidemiology of polymyositis/dermatomyositis. Mt
Sinai J Med 1988;55:447–52.
Pachman LM. Juvenile dermatomyositis: immunogenetics, pathophysiology, and disease expression. Rheum Dis Clin North Am
Counahan R, Winterborn MH, White RH, Heaton JM, Meadow
SR, Bluett NH, et al. Prognosis of Henoch-Schonlein nephritis in
children. Br Med J 1977;2:11–4.
Goldstein AR, White RH, Akuse R, Chantler C. Long-term
follow-up of childhood Henoch-Schonlein nephritis. Lancet 1992;
Stewart M, Savage JM, Bell B, McCord B. Long term renal
prognosis of Henoch-Schonlein purpura in an unselected childhood population. Eur J Pediatr 1988;147:113–5.
Nakamura Y, Yanagawa H, Harada K, Kato H, Kawasaki T.
Mortality among persons with a history of Kawasaki disease in
Japan: the fifth look. Arch Pediatr Adolesc Med 2002;156:162–5.
Besbas N, Ozen S, Saatci U, Topaloglu R, Tinaztepe K, Bakkaloglu A. Renal involvement in polyarteritis nodosa: evaluation of 26
Turkish children. Pediatr Nephrol 2000;14:325–7.
Foeldvari I, Zhavania M, Birdi N, Cuttica RJ, de Oliveira SH,
Dent PB, et al. Favorable outcome in 135 children with systemic
sclerosis: results of a multi-national survey. Rheumatology (Oxford) 2000;39:556–9.
32. Scalapino K, Arkachaisri T, Lucas M, Fertig N, Helfrich DJ,
Londino AV Jr, et al. Childhood onset systemic sclerosis: classification, clinical and serologic features, and survival in comparison
with adult onset disease. J Rheumatol 2006;33:1004–13.
33. Martini G, Vittadello F, Kasapcopur O, Magni Manzoni S, Corona
F, Duarte-Salazar C, et al. Factors affecting survival in juvenile
systemic sclerosis. Rheumatology (Oxford) 2009;48:119–22.
34. Bowyer S, Roettcher P, for the Pediatric Rheumatology Database
Research Group. Pediatric rheumatology clinic populations in the
United States: results of a 3 year survey. J Rheumatol 1996;23:
35. Therneau T, Offord J. Expected survival based on hazard rates
(update): technical report series no. 63. Rochester (MN): Department of Health Sciences Research, Division of Biostatistics, Mayo
Clinic; 1999.
36. Arias E, Anderson RN, Kung HC, Murphy SL, Kochanek KD.
Deaths: final data for 2001. Natl Vital Stat Rep 2003;52:1–115.
37. Hauser TH, Ho KK. Accuracy of on-line databases in determining
vital status. J Clin Epidemiol 2001;54:1267–70.
38. Schisterman EF, Whitcomb BW. Use of the Social Security
Administration Death Master File for ascertainment of mortality
status. Popul Health Metr 2004;2:2.
39. Lin DY, Wei LJ. The robust inference for the proportional
hazards model. J Am Stat Assoc 1989;84:1074–8.
40. Wolfe F, Anderson J, Harkness D, Bennett RM, Caro XJ,
Goldenberg DL, et al. A prospective, longitudinal, multicenter
study of service utilization and costs in fibromyalgia. Arthritis
Rheum 1997;40:1560–70.
41. Macfarlane GL, McBeth J, Silman AJ. Widespread body pain and
mortality: prospective population based study. BMJ 2001;323:
42. Shishehbor MH, Litaker D, Pothier CE, Lauer MS. Association of
socioeconomic status with functional capacity, heart rate recovery,
and all-cause mortality. JAMA 2006;295:784–92.
43. Deal CL, Hooker R, Harrington T, Birnbaum N, Hogan P,
Bouchery E, et al. The United States rheumatology workforce:
supply and demand, 2005–2025. Arthritis Rheum 2007;56:722–9.
44. Linabery AM, Ross JA. Childhood and adolescent cancer survival
in the US by race and ethnicity for the diagnostic period 19751999. Cancer 2008;113:2575–96.
45. Pulte D, Gondos A, Brenner H. Trends in 5- and 10-year survival
after diagnosis with childhood hematologic malignancies in the
United States, 1990-2004. J Natl Cancer Inst 2008;100:1301–9.
46. Nass SJ, Levit LA, Gostin LO, editors. Effect of the HIPAA
privacy rule on health research. In: Beyond the HIPAA privacy
rule: enhancing privacy, improving health through research. Washington (DC): The National Academies Press; 2009. p. 199–244.
47. Lash TL, Silliman RA. A comparison of the National Death Index
and Social Security Administration databases to ascertain vital
status. Epidemiology 2001;12:259–61.
48. Malleson PN, Fung MY, Rosenberg AM. The incidence of pediatric rheumatic disease: results from the Canadian Pediatric
Rheumatology Association Disease Registry. J Rheumatol 1996;
49. Symmons DP, Jones M, Osborne J, Sills J, Southwood TR, Woo P.
Pediatric rheumatology in the United Kingdom: data from the
British Pediatric Rheumatology Group National Diagnostic Register. J Rheumatol 1996;23:1975–80.
Без категории
Размер файла
116 Кб
outcomes, rheumatology, pediatrics, mortality
Пожаловаться на содержимое документа