1.

er Encuentro Entre Neumólogos y Radiólogos,

Madrid, Spain, 2016, October 14th

Difficulties Diagnosing Idiopathic

Pulmonary Fibrosis
Simon Walsh
King’s College Hospital Foundation Trust
London, United Kingdom
 Outline
HRCT pattern revision
Fibrotic lung disease
Radiologic diagnosis of UIP/IPF
Radiologic challenges
Making a HRCT diagnosis in fibrotic lung disease:
key questions….

• Is it a fibrotic lung disease?

• If so, is it classical UIP?

• If no, what are the alternatives?
Making a HRCT diagnosis in fibrotic lung disease:
key questions….

• Is it a fibrotic lung disease?

Revision:
HRCT patterns of fibrosis
(++++ = complete certainty)

• Honeycombing +++(+)

• Traction bronchiectasis ++(+)

• Volume loss +
Revision:
HRCT patterns of fibrosis
(++++ = complete certainty)

• Honeycombing +++(+)

•
How easy is honeycombing to identify?
• 43 Observers
• 27 Expert chest radiologists
• 80 single images

Kw = 0.40 - 0.58
<0.4 = poor, 0.4-0.6 = satisfactory,0.6-0.8 = good,>0.8 = excellent

Watadani et al, Radiology, 2013:266:936

Some inconsistency…..
Weighted kappa’s for honeycombing

• Lynch et al, AJRCCM, 2005:172:488

Kappa = 0.21-0.31

• Sundaram et al, AJR, 2008:191:1032

Kappa = 0.25-0.56

• Goldin et al, Chest, 2009:136:1333

Kappa = 0.39

<0.4 = poor, 0.4-0.6 = satisfactory, 0.6-

0.8 = good, >0.8 = excellent
Some inconsistency…..
Weighted kappa’s for honeycombing

• Lynch et al, AJRCCM, 2005:172:488

Kappa = 0.21-0.31

315 patients with biopsy proven IPF

Expert thoracic radiologists

<0.4 = poor, 0.4-0.6 = satisfactory, 0.6-

0.8 = good, >0.8 = excellent
Identification of honeycombing on
HRCT - cardinal sign of UIP

• False positive identification

– Severe traction bronchiolectasis
– Centrilobular/paraseptal emphysema
• e.g. superimposed on NSIP

– Oedema/infection superimposed on emphysema

– Other cystic conditions e.g. Langerhans CH
Paraseptal Emphysema

Honeycombing
Paraseptal Emphysema

Honeycombing
Paraseptal Emphysema

Honeycombing
Lung biopsy: Fibrotic NSIP and centrilobular emphysema
Revision:
HRCT patterns of fibrosis
(++++ = complete certainty)

• Traction bronchiectasis ++(+)

•
Traction
Bronchiectasis
 Traction
Bronchiectasis

Non-tapering airway
surrounded by
abnormal lung
parenchyma
Identification of traction
bronchiectasis on HRCT
• False positive identification
– Within honeycombing
– Dilated bronchi within OP / DAD
– Conspicuous, but not dilated, bronchi within GGO

• “False negative”
– Within honeycombing (advanced)
– Severity of traction reduced if coexistent emphysema
How easy is traction bronchiectasis to identify?
Observer agreement for traction
bronchiectasis in various FLD
• Fibrotic IIPs (UIP and NSIP)
– Edey 2011 Eur Radiol

• Rheumatoid Arthritis-related FLD

– Kim 2010 Eur Respir J

• Chronic hypersensitivity pneumonitis

– Walsh SLF 2012 Eur Radiol

• All comers connective tissue disease FLD

– Walsh SLF 2014 Thorax

Kappas for traction bronchiectasis = 0.58-0.69

Traction bronchiectasis - identification by
CALIPER software:

…differentiation from honeycombing unreliable

Revision:
HRCT patterns of fibrosis
(++++ = complete certainty)

• Volume loss +
Volume loss can be difficult to appreciate….
Honeycombing

Traction bronchiectasis

Volume loss
Making a HRCT diagnosis in fibrotic lung disease:
key questions….

• If so, is it classical UIP?

 3 diagnosis categories
• UIP pattern (‘definite’, ‘classic’)
• Possible UIP pattern
• Inconsistent with a UIP pattern

Raghu et al. AJRCCM 2011;183:788

Subpleural, basal honeycombing = ‘classic UIP’
Characteristic/Definite UIP
pattern on CT

Supportive CT features (unofficial)…

– “Propeller blade” cranio-caudal distribution
– Nodular ossifications within fibrosis
– Asymmetric distribution of fibrosis
– Component of pleuroparenchymal fibroelastosis
(PPFE)
Supportive/ancillary features of UIP on CT:

• Propeller blade distribution*

– Subpleural disease anterior in upper lobes
– Subpleural disease posterior in lower lobes
Supportive/ancillary features of UIP on CT:
• R to L asymmetry of fibrosis [anecdotal]
Nodular ossification (white spots) in fibrosis
29% prevalence (c.f. 8% in other FLD) [in press]
 Supportive/ancillary features of UIP on CT:

Pleuroparenchymal
Fibroelastosis (PPFE)
+
UIP

PPFE associated with lower zone UIP

in 32% of patients [unpublished data]
 Accuracy of a CT diagnosis of UIP/IPF
Correctness of
Study confident first
choice CT
diagnosis -
‘definite UIP’

JR Mathieson et al, 95%

Radiology, 1989;171:111

KS Lee et al, 100%

Radiology 1994;191:669

S Swenson et al, 100%

Radiology, 1997;205:229

GW Hunninghake, 96%
AJRCCM, 2001;164:193
 Accuracy of a CT diagnosis of UIP/IPF
Correctness of % of cases of UIP
Study confident first without a
choice CT confident CT
diagnosis - diagnosis -
‘definite UIP’ ‘atypical UIP’

JR Mathieson et al, 95% 38%

Radiology, 1989;171:111

KS Lee et al, 100% 39%

Radiology 1994;191:669

S Swenson et al, 100% 33%

Radiology, 1997;205:229

GW Hunninghake, 96% 52%

AJRCCM, 2001;164:193
Definite UIP 32.9%

Probable UIP or possible UIP 61.9%

Possible UIP 4.9%

Richeldi et al, NEJM, 2011;370:2071

What do these atypical cases look like?
Spectrum of atypical radiologic
appearances of biopsy proven UIP

Most common radiologic diagnoses in 34 patients with biopsy

proven UIP whose CT does not meet radiologic criteria for
definite UIP (i.e. basal, subpleural honeycombing):

• NSIP 18
• CHP 4
• Sarcoidosis 3
• OP 1
• Other 8

Sverzellati et al. Radiology 2009;254:987

Recap: Radiologic presentation of UIP/IPF

≈ 30-40% ≈ 60-70%
“Definite UIP” “Not definite UIP”
Radiologic presentation:
Radiologic presentation:
Basal, subpleural,
honeycombing Fibrotic NSIP (most common)*
CHP (rare)
Fibrotic Sarcoidosis (very rare)

≈ 95+% accurate for

histopathologic UIP

*Not entirely clear what “fibrotic NSIP” is on HRCT

Recap: Radiologic presentation of UIP/IPF

≈ 30-40% ≈ 60-70%
“Definite UIP” “Not definite UIP”
Radiologic presentation:
Radiologic presentation:
Basal, subpleural,
honeycombing Fibrotic NSIP (most common)*
CHP (rare)
Fibrotic Sarcoidosis (very rare)

≈ 95+% accurate for

histopathologic UIP

*Not entirely clear what “fibrotic NSIP” is on HRCT

Some pitfalls:
IMPORTANT: confidence ≠ accuracy
Fiducia ≠ precisione
RADIOLOGIST MDT DIAGNOSIS
Chronic
IPF hypersensitivity
pneumonitis

IPF Fibrotic NSIP

Sarcoidosis
IPF (very rare)
 HRCT diagnosis fibrotic lung disease

• If no, what are the alternatives?

Fell et al. AJRCCM, 2010;181:832-837
Column 3 as a checklist to differentiate UIP
from others fibrosing lung diseases

• Not basal
Chronic
• Bronchocentric hypersensitivity
• Mosaicism (lobules) pneumonitis

• Nodules
• GGO > reticular
• Cysts
• Consolidation
 OP
UIP
NSIP LIP

HP
DIP/
RB-ILD

pre 2002 ATS/ERS IIP classification

 OP

UIP
NSIP LIP

HP
DIP/
RB-ILD

Silva et al, Radiology 2008;246:288

 OP

UIP
NSIP LIP

HP
DIP/
RB-ILD

2016…
HRCT pointers to chronic hypersensitivity
pneumonitis:
• Lobules of decreased attenuation in spared (non-
fibrotic) lung

• Occasional septal thickening may be a bit more

obvious than in other fibrotic IIPs

• x3 distributions of fibrosis: UZ, LZ or random

- sometimes vague/subtle bronchocentricity if UZ

• Coexistent subacute changes - indistinct relatively

low attenuation centrilobular nodules (rare)
Lobules of decreased attenuation in spared lung
Chronic HP

UIP
Septal thickening in chronic
hypersensitivity pneumonitis
Unusual distribution of fibrosis, particularly vague
bronchocentricity when upper lobe predominant:

n.b. Bronchocentricity, when present,

is much more subtle than the
bronchocentric fibrosis in sarcoidosis
Coexisting chronic and subacute features
of HP (relatively rare)
 BAL lymphocytosis 27% UIP on lung biopsy
MDT diagnosis of Chronic HP MDT diagnosis of IPF
Column 3 as a checklist to differentiate UIP
from others fibrosing lung diseases

• Not basal
• Bronchocentric
• Mosaicism (lobules)
• Nodules
• GGO > reticular
• Cysts
• Consolidation
 DMH

SLFW AD NS
 Inconsistent

All could pass for “possible UIP”

Inconsistent
Inconsistent
Biopsy == UIP
IPF Scleroderma

IPF Smoking related ILD (RBILD/DIP)

Let’s get back to IPF…
Radiologic diagnosis:
The IPF challenge

• Misdiagnosis of IPF
• Clarify status of “possible UIP”
• Detection of early disease
ATS/ERS/JRS/ALAT guidelines:
Radiologic misdiagnosis is a problem
Clinicians perspective:

“…majority of clinicians found a HRCT diagnosis of IPF to be

reliable in less than a third of practicing radiologists…..”

AU Wells - straw poll,

Berlin IPF AIR meeting, November 2011
ESTI (European Society of Thoracic Imaging)

BSTI (British Society of Thoracic Imaging)

STR (Society of Thoracic Radiology)

KSTR (Korean Society of Thoracic Radiology)

 Observer agreement:
112 thoracic radiologists from 5 continents

Kw = 0·47 ± 0.05

<0.4 = poor,
0.4-0.6 = satisfactory,
0.6-0.8 = good,
>0.8 = excellent
Walsh et al, Thorax 2016;71:45-51
 Observer agreement:
> 25 years experience (n=22)

Kw = 0·37 ± 0.11

<0.4 = poor,
0.4-0.6 = satisfactory,
0.6-0.8 = good,
>0.8 = excellent
Walsh et al, Thorax 2016;71:45-51
Example
112 thoracic radiologists
ESTI, STR, BSTI, KSTR

Honeycombing - 47.8%
UIP – 33.9%
Possible UIP – 56.5%

Walsh et al, Thorax 2016;71:45-51

112 thoracic radiologists
ESTI, STR, BSTI, KSTR
112 thoracic radiologists
ESTI, STR, BSTI, KSTR

Deciding between the presence of

traction bronchiectasis OR honeycombing
can be the difference between
“definite UIP” and “possible UIP”
Radiologist’s Challenge 2:
Clarify “possible UIP” status

• Significant proportion of IPF patients

have HRCT appearances that meet
“possible UIP” criteria

• Many unable or unwilling to undergo

lung biopsy

• “Guideline-based” IPF diagnosis not

possible
Radiologist’s Challenge 2:
Clarify “possible UIP” status

Are there specific CT features that

predict histopathologic UIP?

Fell et al. AJRCCM, 2010;181:832-837

Clinical predictors of IPF
The importance of age

IPF or fibrotic NSIP?

Fell et al. AJRCCM, 2010;181:832-837
Clinical predictors of IPF
The importance of age

In idiopathic fibrotic lung disease,

increasing age increases likelihood
of a diagnosis of UIP/IPF

>70 years – PPV 95%

>75 years – PPV 100%

Fell et al. AJRCCM, 2010;181:832-837

• Age >= 60
• Reticular abnormality > 1/3 total lung
– Probability of IPF >80%
– Specificity for IPF 96%
– 16% of biopsies avoided

Salisbury et al. Respir Med 2016 Sep;118:88-9

Gruden et al 2013

Raghu et al 2014

Chung et al 2015
 ‘Coarse heterogeneous fibrosis without
honeycombing’ highly predictive of histologic UIP
and a UIP-like disease course

Gruden et al, AJR 2013;200:458

Counter argument:
How important is this distinction?

Inpulsis® HRCT entry criteria

== “Possible UIP”
(with traction bronchiectasis)

Richeldi et al, NEJM, 2014;370:2071-82

Definite UIP 32.9%

Probable UIP or possible UIP 61.9%

Possible UIP 4.9%

i.e. honeycombing present in only 1/3 patients,

in practice may be 20-45%

Richeldi et al, NEJM, 2011;370:2071

Best use of HRCT?
“Working diagnosis of IPF”*

• Less dependent on absence/presence of specific

HRCT patterns
– HRCT ≠ “truth data” when histopathology unavailable
(e.g. note importance of age*)

• Take CT appearance as starting point

– Dynamic diagnostic process
– Integrate other important non-radiological information,
in particular DISEASE BEHAVIOUR

*Athol Wells, personal communication

*CD Fell et al, AJRCCM 2010;181:832
 An evaluation of interobserver agreement
and accuracy for “solo” clinicians diagnoses
in the setting of diffuse lung disease

• 60 cases of DPLD
• Provide a differential diagnosis
• 673 clinician's signed up (21/9/2016)
• 11,737 patient evaluations (21/9/2016)

www.theipfproject.com
wKappa’s for probability of an IPF diagnosis

Overall (n=125) = 0.55

Expert panel (n=14) = 0.68
No MDT available (n=25) = 0.49
“Dedicated” ILD MDT (n=65) = 0.58

www.theipfproject.com
 Inter-MDT agreement for diagnostic
likelihood of a diagnosis of IPF

Kw = 0.71
<0.4 = poor,
0.4-0.6 = satisfactory,
0.6-0.8 = good,
>0.8 = excellent
Walsh SLF et al. Lancet RM, 2016;4:557-65
 Conclusion
• Radiologic diagnosis of UIP/IPF
– “Working diagnosis of IPF”* approach
– HRCT as a starting point

• Challenges in 2016 for the radiologist

– Agreement issues for key observations
• Possible and inconsistent with UIP categories