Henoch Schnlein Purpura: A role for corticosteroids during hospitalization

1
Henoch Schönlein PurpuraA role for
corticosteroids during hospitalization

• Pamela Weiss, MD
• Pediatric Rheumatology Fellow
• Childrens Hospital of Philadelphia
• Center for Pediatric Clinical Effectiveness
  Seminar Series
• November, 2007

2
Outline

• Review the literature on corticosteroids and HSP
• Describe CS use for HSP during hospitalization
• Compare the outcomes of CS-treated and untreated
  hospitalized children with HSP
• Describe challenges in analysis with confounding
  by indication

3
HSP Whats in a name?
Description gently periodic sticking pains in
the joints which are odematous, swollen and
very painful the characteristic spots of the
disease appear in the majority of cases first on
the extremities and particularly on the lower
onesthe spots are small, the size of a lentil.
The fever disappears through the skin and urine
crises, but the exanthema usually remains after
the crisis
Johann Lukas Schönlein (1793-1864)
4
HSP Whats in a name?
Description gently periodic sticking pains in
the joints which are odematous, swollen and
very painful the characteristic spots of the
disease appear in the majority of cases first on
the extremities and particularly on the lower
onesthe spots are small, the size of a lentil.
The fever disappears through the skin and urine
crises, but the exanthema usually remains after
the crisis
Johann Lukas Schönlein (1793-1864)
5
HSP Whats in a name?
Description a 15 year old boy who.developed
a gastroduodenal catarrh with a slight icterus. A
few days later, pain developed in the finger
joints of both hands without any swelling. After
a few days an extensive purpura appeared on the
abdomen and skin of the thighs, and very soon
several intestinal symptoms appeared, intense
colic, vomiting and very black stools. ..In all
5 attacks took place, which with the intervals,
occupied a period of 7 weeks.
Eduard Henoch (1820-1910)
6
HSP Whats in a name?
Description a 15 year old boy who.developed
a gastroduodenal catarrh with a slight icterus. A
few days later, pain developed in the finger
joints of both hands without any swelling. After
a few days an extensive purpura appeared on the
abdomen and skin of the thighs, and very soon
several intestinal symptoms appeared, intense
colic, vomiting and very black stools. ..In all
5 attacks took place, which with the intervals,
occupied a period of 7 weeks.
Eduard Henoch (1820-1910)
7
HSP therapy have we come far?
1800
1850
1900
1950
2000
X
X
X

• 1950s
• Corticosteroids?
• ACTH

• 1837 (Schönlein)
• Cream of tartar water
• Ammonium acetate
• Dandelion extract with rhubarb tartar
• Calamus with ammonia
• Ammonium succinate
• Bitter tea with car, bened, or vermouth

• 1874 (Henoch)
• Opium
• Emulsion of sweet almond oil and milk
• Castor oil
• Chloral hydrate
• Calomel
• Ice to the abdomen

2007 We still dont know how to treat HSP
8
HSP a quick look

• Palpable purpura in the absence of
  thrombocytopenia and coagulopathy
• Plus one or more of the following
• Diffuse abdominal pain
• Arthritis or arthralgia
• Biopsy with IgA deposition
• EULAR/ PReS consensus criteria. 2006

9
Why study HSP?

• Most common vasculitis of childhood
• 8-20/100,000 children annually
• 49 of US childhood vasculitides
• Chronic renal failure
• No consensus for therapy
• Few experimental studies

Granular mesangial pattern of IgA
immunofluorescence
10
HSP - self limited?

• Only if you dont mind suffering for 2-3 weeks
• Renal involvement in 60
• 20 may develop RPGN
• 40 - 70 of pregnancies complicated by
• Hypertension
• Proteinuria
• Pre-eclampsia

Levy et al. Adv Nephrol Necker Hosp
1976. Ronkainen et al. Clin Nephrol 2003.
Ronkainen et al. Lancet 2002. Goldstein et al.
Lancet 1992.
Scarred glomerulus
11
Outline

• Review the literature on corticosteroids and HSP
• Describe CS use for HSP during hospitalization
• Compare the outcomes of CS-treated and untreated
  hospitalized children with HSP
• Describe challenges in analysis with confounding
  by indication

12
Meta-analysis HSP and CS efficacy

• For children with HSP, do CS decrease
• Renal sequelae?
• Surgical intervention?
• Duration of abdominal pain?
• Recurrence?

13
Reasons for Exclusion of Articles
14
Papers included
15
CS and odds of persistent renal disease
Reduced odds of developing persistent renal
disease
16
CS and odds of surgical intervention
Suggest reduced odds of surgery
17
CS and abdominal pain resolution
Reduced mean resolution time of abdominal pain
Increased odds of pain resolution within 24h
18
CS and odds of recurrence
Suggest reduced odds of recurrence
19
What study would reverse the findings?
20
Meta-analysis conclusions

• Results suggest CS are helpful for
• Prevention of persistent renal disease
• Abdominal pain
• Surgery
• Recurrence

21
Outline

• Review the literature on corticosteroids and HSP
• Describe CS use for HSP during hospitalization
• Compare the outcomes of CS-treated and untreated
  hospitalized children with HSP
• Describe challenges in analysis with confounding
  by indication

22
Pediatric Health Information System (PHIS)

• Administrative database
• Inpatient data
• 46 pediatric hospitals
• Administrative data
• Demographics
• Admission and discharge diagnoses
• Billing
• De-identified data
• Reliability and validity checks

23
How are we using CS for inpatient HSP?

• Objective To describe the clinical management of
  HSP
• Variation in the use of CS among hospitals
• Use and timing of CS
• Methods
• Retrospective cohort study from 2000-2006 using
  PHIS
• Cohort subjects with a discharge HSP ICD-9-CM
• Exposure pharmacy billing data
• Receipt of CS on hospital day 1
• Analysis Conditional logistic regression
• Adjustment for clustering within hospitals
• Adjustment for admission diagnosis

24
Patients
Admission severity scores available for 1,458
1st admissions and 316 readmissions. As
indicated by admission ICD-9-CM codes. ICD-9-CM
related to nephritis, hematuria, or proteinuria.
25
Variation in the use of CS among hospitals
1
.8
.6
Proportion of cases that received CS on hospital
day 1
.4
.2
0
0
10
20
30
Hospital volume of HSP cases/year during study
period
26
Factors associated with CS use on hospital day
1

• Admission categories HSP, gastrointestinal
  bleeding/abdominal symptoms, renal involvement/
  hypertension, arthritis/arthralgias/limb pain,
  intussusception, rash, orchitis, myalgias/ fever/
  non-specific arteritis, missing admission
  diagnosis
• Age categories lt3y, 3-4y, 5-6y, 7-11y, gt12y
• Includes pain, hepatitis, non-infectious
  gastroenteritis, vomiting alone, pancreatitis,
  appendicitis, intussusception
• Includes hematuria, proteinuria, renal failure

27
Do the significant predictors differ if we expand
the definition of CS exposure?
28
Factors associated with CS use during
readmission
29
In summary
What matters is. Admission hospital Admission
diagnosis Age Discharge diagnoses in addition
to HSP
30
Outline

• Review the literature on corticosteroids and HSP
• Describe CS use for HSP during hospitalization
• Compare the outcomes of CS-treated and untreated
  hospitalized children with HSP
• Describe challenges in analysis with confounding
  by indication

31
Does early CS use during hospitalization affect
HSP outcomes?

• Objectives To compare HSP outcomes of children
  who receive corticosteroid vs no CS
• Primary outcome Length of hospitalization
• Secondary outcomes
• Use of parenteral nutrition
• Admission to the intensive care unit (ICU)
• Surgery (abdominal, open biopsy, or arthroscopy)
• Abdominal imaging or endoscopy
• Renal biopsy
• Side effects from CS

32
Methods

• Data source PHIS
• A total of 38 hospitals and 200 years of data
• Subjects
• Inclusion criteria
• lt18 years of age with an ICD-9-CM diagnosis
  indicating HSP (code 287.0)
• Admission between Jan 1, 2000 Dec 31, 2006
• Exclusion criteria
• Any HSP in a 6 month look-back period
• Diagnosis of another rheumatic condition
  (Wegeners granulomatosis, Lupus, or
  Polyarteritis nodosa)

33
Methods

• Exposure
• Pharmacy billing data indicating administration
  of CS on hospital day 1
• Outcomes
• Length of stay
• Use of parenteral nutrition
• Admission to the intensive care unit (ICU)
• Surgery (abdominal, open biopsy, or arthroscopy)
• Abdominal imaging
• Endoscopy
• Renal biopsy

34
Methods

• Analysis
• Multivariate logistic and linear regression
  models
• Adjusted for pre-specified covariates
• Age, sex, race, abdominal pain, gastrointestinal
  bleeding, hypertension, urinary abnormalities,
  hospital-level mean CS-use, early vs delayed HSP
  diagnosis, discharge month and year
• Clustering within hospitals

35
Outcomes
Shorter length of stay by 0.93 days (95 CI
-1.37, -0.49 plt0.01)
36
Sensitivity analysis CS-exposure definition
37
Readmissions

• 272 children (15) had gt1 or more (range 1-9)
• Median time to 1st readmission 7 days (range
  1- 838)
• 90 occurred within 90 days
• Higher severity than initial admissions
• CS did not reduce the odds of readmission
• OR 1.23 (95 CI 0.91, 1.67)

38
Is there a CS dose-response?
Yes!
39
Side effects from CS

• 16 (1) during initial hospitalization
• 6 (1.5) during readmission
• The CS adverse event was identifiable as
• Hallucinations (2)
• Hyperglycemia or glucosuria (2)
• Anxiety (1)

40
Limitations

• Outpatient data not captured
• Only capture ICD-9-CM coded diagnoses
• Potential for
• Misdiagnosis by clinicians
• Miscoding
• Cant capture mild recurrences

41
Outline

• Review the literature on corticosteroids and HSP
• Describe CS use for HSP during hospitalization
• Compare the outcomes of CS-treated and untreated
  hospitalized children with HSP
• Describe challenges in analysis with confounding
  by indication

42
Association
43
Confounding
44
Confounding by Indication
45
Confounding by Indication
46
Confounding by Indication
47
Propensity Score Matching
Breaks the association between treatment
likelihood and severity of condition
48
Propensity Score Matching
Breaks the association between treatment
likelihood and severity of condition but does
so reliably only if the association is similar
across sites of care
49
Heterogenous Confounding by Indication

• Unlike other sources of confounding, we have some
  insights about the mechanisms by which
  confounding by indication occurs, namely
  prescribing health care professionals have views
  about
• The efficacy of a drug
• The safety or cost of a drug
• The prognosis for the condition
• These views, in turn, create a range of
  severity--treatment-likelihood associations

50
Heterogenous Confounding by Indication

• We hypothesize that the pattern of the
  severity--treatment-likelihood association varies
  from hospital-to-hospital, based on
• Evidence from this study
• Sociology of thought leaders
• Clinical experience
• This hypothesis encourages the main analysis to
  be conditional, in that the severity--drug-exposur
  e will vary by site, so the drug--outcome would,
  too.

51
Heterogenous Confounding by Indication

• Conditional Propensity Score
• Score estimated for each site
• Matching constrained to occur only within sites

52
Heterogenous Confounding by Indication

• Conditional multivariable regression
• Each site treated as a group
• All relevant individual parameters
  included in the model
• Summary parameter for each site included in the
  model (in this case, the overall proportion of
  patients who received treatment, ie, the mean
  treatment exposure
• Can examine whether this summary parameter
  interacts with other individual level parameters

53
Acknowledgements
Chris Feudtner Russell Localio Sandy Burnham Ron
Keren Kari Hexem
NIH NIDDK F32