introduction to the design of diagnostic criteria

1
introduction to the design of diagnostic criteria
Joop P van de MerweDept. of Immunology
Internal MedicineErasmus MC Rotterdamj.vandemerw
e_at_erasmusmc.nl ESSIC Meeting Baden 16-18 June
2005
2

Fries JF et al. Arthritis Rheum 199437454-62
3
CLASSIFICATION CRITERIA
Classification criteria separate patients with
the disease from the general population and from
patients with potentially confusable conditions
Fries JF et al. Arthritis Rheum 199437454-62
(modified)
4
the problem
overlapping features example systemic
autoimmune diseases
5
systemic autoimmune diseases (1)
examples

• rheumatoïd arthritis
• Sjögren's syndrome
• systemic lupus erythematosus (SLE)
• antiphospholipid syndrome (APS)
• mixed connective tissue disease (MCTD)
• systemic sclerosis (scleroderma, CREST-syndrome)
  
• polymyositis / dermatomyositis
• relapsing polychondritis

6
systemic autoimmune diseases (2)
possible features

• clinical
• arthritis / arthralgia
• vasculitis
• fever
• Raynaud's phenomenon
• fatigue
• various renal diseases

• laboratory
• increased ESR, CRP
• anaemia
• leukopenia
• rheumatoid factor
• ANA
• trombopenia

not specific for one single disease
not specific for one single disease
7
systemic autoimmune diseases (3)
possible accompaning organspecific autoimmune
diseases

• atrophic gastritis - pernicious anaemia
• Hashimoto's disease
• keratoconjunctivitis sicca
• lymfocytic interstitial pneumonitis
• myositis
• pericarditis, pleuritis
• peripheral neuropathy
• uveitis (iritis, iridocyclitis)
• scleritis
• interstitial cystitis

no combination is specific for one single disease
8
systemic autoimmune diseases (4)
how do you recognize them ?
9
people
how do you recognize them ?
each person has a face
10
people
how do you recognize them ?
none of the parts of the face is unique, but
the combination is
11
people
how do you recognize them ?
none of the parts of the code is unique, but
the combination is
12
systemic autoimmune diseases (4)
how do you recognize them ?
each autoimmune disease has its own "face"
13
systemic autoimmune diseases (5)
characteristic features
characteristic feature specific way it affects
the lacrimal and salivary glands specific way
it affects the joints specific way it affects
the skin specific way it affects the skin
combination of clinical symptoms and
anti-RNP specific way it affects the skin
combination of symptoms
disease Sjögren's syndrome rheumatoid
arthritis systemic lupus erythematosus subacute
cutaneous lupus erythematosus mixed
connective tissue disease (MCTD) systemic
sclerosis (scleroderma) CREST syndrome
14
characteristic features
what are the characteristic features of IC ?

• what is the consequence of the PBS/IC concept ?

ICS definition of PBS/IC PBS "suprapubic pain
related to bladder filling, accompanied by other
symptoms such as increased daytime and nighttime
frequency in the absence of infection or other
pathology" IC above with "typical cystoscopic
and histological features"
15
characteristic features
what are the characteristic features of IC ?
ICS definition of IC may be written
as suprapubic pain related to bladder filling,
accompanied by other symptoms such as increased
daytime and nighttime frequency with typical
cystoscopic and histological features in the
absence of infection or other pathology
what are the "typical cystoscopic and
histological features of IC" ? what has happened
to urgency ?
16
characteristic features
what are the "typical cystoscopic and
histological features of IC" ?

• Cysto features Hunner's ulcer or low capacity
• epithelial denudation, submucosal inflammation,
• granulation tissue, edema, congestion,
• haemorrhage detrusor fibrosis (and myopathy)
• increased mast cell number activation and
• neuronal staining

from Hanno P, Burks D. Painful bladder syndrome
/ interstitial cystitis. Course 93 IC, AUA 2005
Annual Meeting, San Antonio, TX, May 21-26,2005
17
characteristic features

• Cysto features Hunner's ulcer or low capacity
• epithelial denudation, submucosal inflammation,
• granulation tissue, edema, congestion,
• haemorrhage detrusor fibrosis (and myopathy)
• increased mast cell number activation and
• neuronal staining

but these are not unique for IC !
from Hanno P, Burks D. Painful bladder syndrome
/ interstitial cystitis. Course 93 IC, AUA 2005
Annual Meeting, San Antonio, TX, May 21-26,2005
does this matter as long as the combination is
unique ?
18
PBS / IC concept
PBS
19
gold standard
gold standard diagnosis by experts why ? - the
diagnosis by experts is the best standard we
have as long as the cause of a disease is
unknown advantage - non-experts can diagnose a
disease as if they were experts (skill level) -
diagnoses are made on the same basis by experts
and non-experts (uniformity) - rules are based on
information from many experts (acceptance and
support)
20
classification methods (1)
classification tree the classification tree is
constructed by repeated splits of groups into
2 descendant subgroups
Arthritis Rheum 1990331137-44
21
classification tree

• pro
• items may be polychotomous or continuous
• nonparametric no reference to a model for the
• relationship between classification items and
  disease
• status
• high information content - classified groups of
  subjects in
• studies may be referred to the exact
  classifying subgroup
• of the tree

Arthritis Rheum 1990331137-44
22
"number of criteria present rule"
Ann Rheum Dis 200261554-8
23
classification methods (2)
"number of items present rule"

• pro
• simple definitions (/-) easy application

• con
• simple definitions (/-)
• - all items are given equal weight
• - all items must be dichotomous (/-)
• ? little insight into the characteristics of a
• classified patient group

Arthritis Rheum 1990331137-44
24
suggested approach in a multicentre study
1. collect what information (symptoms and signs)
is used by experts to diagnose IC and
differentiate IC from confusable diseases
for both men and women
requires definitions and decisions on data
format
25
suggested approach in a multicentre study
2. collection of data from patients with IC
and patients with confusable diseases to
define classification criteria for IC
(training phase) sensitivity
specificity
calculate an optimal combination on the basis of
how many false negative and false
positive diagnoses we accept
26
suggested approach in a multicentre study
3. validation with new patient groups
the same database can be used to create number
of items present rule for clinical diagnoses and
a classification tree for scientific studies
27
data format
data may be - continuous avoid entry of
cut-off points for continuous data but have
the statistical analysis decide - categorical
28
to be evaluated .
tests to be evaluated for their diagnostic
value should be done in all patients with IC and
confusable diseases 1. antiproliferative
factor 2. ................................ 3.
................................
29
summary conclusions (1)
1. decision on broad definition of IC PBS
"typical" cystoscopy and/or histology 2. select
diseases with which IC can be confused
(confusable diseases, CD) list to be
discussed this morning 3. what information is
needed to distinguish IC from CD follows
from (2) but needs discussion and decision 4.
decision on what percentage of false negative
and false positive diagnoses we accept
30
summary conclusions (2)
5. decision on gold standard for the diagnosis
IC diagnosis by experts 6. decision on methods
of classification classification tree and
"number of items present rule" 7. decision
to perform a small pilot to test
procedures 8. decision on tests to be evaluated
APF test
31
consensus on IC