; Nonspecific Interstitial Pneumonia and Idiopathic Pulmonary Fibrosis
Learning Center
Plans & pricing Sign in
Sign Out
Your Federal Quarterly Tax Payments are due April 15th Get Help Now >>

Nonspecific Interstitial Pneumonia and Idiopathic Pulmonary Fibrosis


  • pg 1
									                 Note: This copy is for your personal non-commercial use only. To order presentation-ready
                 copies for distribution to your colleagues or clients, contact us at www.rsna.org/rsnarights.

                                                                                                                                                ORIGINAL RESEARCH
                                                             Nonspecific Interstitial
                                                             Pneumonia and Idiopathic
                                                             Pulmonary Fibrosis: Changes in

                                                                                                                                               THORACIC IMAGING
                                                             Pattern and Distribution of Disease
                                                             over Time1
C. Isabela S. Silva, MD, PhD
                                                                  Purpose:    To retrospectively assess the change in disease pattern of
Nestor L. Muller, MD, PhD
                                                                              nonspecific interstitial pneumonia (NSIP) and idiopathic
David M. Hansell, MD
                                                                              pulmonary fibrosis (IPF) findings seen at thin-section com-
Kyung S. Lee, MD                                                              puted tomography (CT) at long-term follow-up and to com-
Andrew G. Nicholson, MD                                                       pare the same with initial findings at CT.
Athol U. Wells, MD
                                                              Materials and   The study included 48 patients (28 men, 20 women; mean
                                                                  Methods:    age, 57.5 years) with biopsy-proved NSIP (n 23) or IPF
                                                                              (n     25) who underwent CT at initial diagnosis and at
                                                                              follow-up 34 –155 months later. The CT scans were ran-
                                                                              domized and reviewed by two independent thoracic radiol-
                                                                              ogists for pattern and distribution of ground-glass opacity
                                                                              (GGO), reticulation, traction bronchiectasis and bronchi-
                                                                              olectasis, and honeycombing. Statistical analysis was per-
                                                                              formed by using nonparametric methods and univariate
                                                                              logistic regression.

                                                                   Results:   Follow-up CT in patients with NSIP showed marked de-
                                                                              crease in the extent of GGO, increase in reticulation, and a
                                                                              greater likelihood of peripheral distribution (all P .05).
                                                                              At presentation, the CT findings were interpreted as sug-
                                                                              gestive of NSIP in 18 of 23 patients with NSIP and indeter-
                                                                              minate or suggestive of IPF in five. In five (28%) of 18
                                                                              patients with initial findings suggestive of NSIP, the fol-
                                                                              low-up CT scans were interpreted as more suggestive of
                                                                              IPF. No CT features seen at presentation allowed distinc-
                                                                              tion between patients with NSIP that maintained an NSIP
                                                                              pattern at follow-up and those that progressed to an IPF

                                                                Conclusion:   At follow-up CT, 28% of patients with initial CT findings
                                                                              suggestive of NSIP progressed to findings suggestive of IPF.
                                                                              Similar initial CT findings for NSIP may have different
  From the Department of Radiology, Vancouver General                         imaging outcomes.
Hospital, University of British Columbia, 3350-950 W 10th
Ave, Vancouver, BC, Canada V5Z 4E3 (C.I.S.S., N.L.M.);                         RSNA, 2008
Department of Radiology (D.M.H.), Department of Pathol-
ogy (A.G.N.), and Interstitial Lung Disease Unit (A.U.W.),
Royal Brompton Hospital, London, England; and Depart-
ment of Radiology and the Center for Imaging Science,
Samsung Medical Center, Sungkyunkwan University
School of Medicine, Seoul, Korea (K.S.L.). Received Feb-
ruary 22, 2007; revision requested May 8; revision re-
ceived June 12; accepted July 18; final version accepted
September 11. Address correspondence to C.I.S.S.
(e-mail: isabela.silva@vch.ca ).

    RSNA, 2008

Radiology: Volume 247: Number 1—April 2008                                                                                               251
THORACIC IMAGING: Nonspecific Interstitial Pneumonia: Changes over Time                                                                           Silva et al

       he idiopathic interstitial pneumo-        mimic those of other interstitial lung      surgical lung biopsy at one of three large
       nias are a group of interstitial lung     diseases, particularly NSIP. The most       teaching hospitals between 1993 and
       diseases characterized by varying         common thin-section CT manifestation        2005. Only those patients with a final
patterns of inflammation and fibrosis              of NSIP consists of symmetric bilateral     clinical diagnosis of idiopathic UIP (IPF)
(1). The two most common forms are               areas of ground-glass opacity (GGO)         and idiopathic NSIP, who underwent
idiopathic pulmonary fibrosis (IPF),              with superimposed fine reticular opaci-      thin-section CT at the time of initial di-
characterized by a histologic pattern of         ties, often with traction bronchiectasis    agnosis and underwent follow-up CT a
usual interstitial pneumonia (UIP), ac-          and bronchiolectasis but with no or only    minimum of 3 years later, were in-
counting for up to 65% of patients, and          mild honeycombing (10–14). Therefore,       cluded. Patients with coexistent pneu-
nonspecific interstitial pneumonia                the findings of NSIP, particularly fibrotic   monia and cardiac failure at the initial
(NSIP), which accounts for 4%–36% of             NSIP, may resemble those of IPF (10,        or follow-up studies were excluded. The
patients (2). On average, the outcome is         11,14).                                     patient group (n       48) consisted of 23
substantially better in NSIP than in IPF              Distinction between IPF and NSIP is    patients with NSIP and 25 patients with
(2). However, the 5- and 10-year mor-            important because NSIP has a consider-      IPF (28 men, 20 women; age range,
tality in fibrotic NSIP has been substan-         ably better prognosis than IPF (1). To      32–78 years; mean, 57.5 years          9.3
tial in some series (3,4). It appears that       our knowledge, there is limited infor-      [standard deviation]). The proportions
a sizeable minority of NSIP patients fol-        mation regarding the change in pattern      of male to female patients were statisti-
low an IPF-like disease course, with             and distribution of thin-section CT find-    cally different in NSIP and IPF; there
progression despite treatment and a              ings of NSIP at long term follow-up (15–    was a larger proportion of men in IPF
similar mortality to that seen in IPF pa-        19). Given the presence of a subgroup       (20 of 25, 80%) than in NSIP (eight of
tients (5). It has been suggested that in        of NSIP patients with inexorably pro-       23, 35%) (P .005, 2 analysis). There
some patients, NSIP may represent an             gressive disease (5), we hypothesized       was no significant (P        .10, Student t
early stage of UIP or, alternatively, that       that at least some patients with histo-     test) difference in age between the NSIP
an NSIP pattern seen at biopsy may sim-          logic diagnoses of NSIP at surgical bi-     and IPF groups (mean, 55.7 years 10.5
ply be indicative of relatively inactive         opsy may progress from an NSIP pat-         and 59.1 years 8.1, respectively).
UIP (2,6). However, Katzenstein et al            tern to a UIP pattern at follow-up CT.           The interval between the initial and
(7) compared lung biopsy and subse-              Thus, the purpose of our study was to       follow-up CT scans ranged from 34 –155
quent explant tissues in 20 patients with        retrospectively assess the change in dis-   months (median, 61 months) in the
idiopathic interstitial pneumonias and           ease pattern of NSIP and IPF findings        NSIP patients and 34 –131 months (me-
found no explant showing UIP that was            seen at thin-section CT at long-term fol-   dian 66 months) in the IPF patients. All
preceded by biopsy findings of NSIP.              low-up and to compare the same with         patients underwent treatment with cor-
     The important role of thin-section          initial findings at CT.                      ticosteroids, with or without immuno-
computed tomography (CT) in the eval-                                                        suppressive agents, between the initial
uation of idiopathic interstitial pneumo-                                                    and follow-up CT scans. The dose and
nias is well known (1,8). The character-           Materials and Methods                     duration of treatment was determined
istic CT features of IPF consist of a retic-
ular pattern, traction bronchiectasis,           Patients
and honeycombing that involves mainly            Approval for this study was obtained        Published online before print
the peripheral lung regions and the lung         from the institutional clinical research    10.1148/radiol.2471070369
bases (8). In approximately one-half of          ethics board of the three hospitals par-
                                                                                             Radiology 2008; 247:251–259
patients with IPF, a confident diagnosis          ticipating in the study. Informed con-
can be made on the basis of clinical and         sent was waived.                            Abbreviations:
thin-section CT findings, precluding the               Patient selection was made by a re-    GGO ground-glass opacity
                                                                                             IPF idiopathic pulmonary fibrosis
need to obtain a surgical biopsy (9).            view of the medical records of all pa-
                                                                                             NSIP nonspecific interstitial pneumonia
However, in the remaining 50% of pa-             tients who had received a histologic di-    UIP usual interstitial pneumonia
tients the thin-section CT findings may           agnosis of NSIP or UIP on the basis of
                                                                                             Author contributions:
                                                                                             Guarantors of integrity of entire study, C.I.S.S., N.L.M.;
 Advance in Knowledge                              Implication for Patient Care              study concepts/study design or data acquisition or data
                                                                                             analysis/interpretation, all authors; manuscript drafting or
      Our study shows that at 3 years                The results of this study indicate      manuscript revision for important intellectual content, all
      or longer follow-up, 28% of pa-                that there are no CT features at        authors; manuscript final version approval, all authors;
      tients with initial CT findings sug-            presentation that allow distinction     literature research, C.I.S.S., N.L.M., K.S.L., A.U.W.; clinical
      gestive of nonspecific interstitial             between patients with NSIP that         studies, C.I.S.S., N.L.M., D.M.H., K.S.L., A.G.N.; statistical
      pneumonia (NSIP) progress to                   maintain an NSIP pattern from           analysis, C.I.S.S., N.L.M., A.U.W.; and manuscript editing,
                                                                                             all authors
      findings suggestive of idiopathic               those that progress to an IPF pat-
      pulmonary fibrosis (IPF).                       tern at follow-up.                      Authors stated no financial relationship to disclose.

252                                                                                                       Radiology: Volume 247: Number 1—April 2008
THORACIC IMAGING: Nonspecific Interstitial Pneumonia: Changes over Time                                                                                      Silva et al

by the clinical conditions and varied be-        tients, the images were reviewed on a                     predominance of abnormalities along
tween the three institutions participat-         dedicated picture archiving and commu-                    the bronchi and vessels), peripheral (a
ing in the study.                                nication system workstation (CS 5000,                     predominance of abnormalities in the
                                                 AGFA; Mortsel, Belgium).                                  outer one-third of the lung or along
Histologic Diagnosis                                                                                       the interlobar fissures), or random
In all patients, the histologic diagnoses        Image Evaluation                                          (no peribronchovascular or peripheral
of UIP and NSIP were made by consen-             The thin-section CT scans (initial and                    predominance). Zonal predominance
sus agreement between at least two pa-           follow-up) were randomized and re-                        of abnormalities was assessed as being
thologists (6 –21 years experience in            viewed by two independent fellow-                         upper, lower, or random. Upper lung
lung pathology) by using the combined            ship-trained thoracic radiologists                        zone predominance was considered
American Thoracic Society and Euro-              (N.L.M. and C.I.S.S., with 22 and 4                       present when disease extent was
pean Respiratory Society Consensus               years experience, respectively) with-                     greatest above the level of the tracheal
Classification criteria (1). The histologic       out knowledge of clinical information                     carina; lower zone predominance was
diagnosis of UIP was made given the              or histologic diagnosis. The readers                      considered present when disease ex-
presence of temporal heterogeneity               were aware that only patients with                        tent was greatest below this level. Rel-
with nonuniform and variable intersti-           NSIP and IPF were included in the                         ative sparing of the lung immediately
tial changes, including intermingled             study, but did not know the diagnosis                     adjacent to the pleura in the dorsal
zones of established interstitial fibrosis,       or whether it was the initial or fol-                     regions of the lower lobes (relative
inflammation, fibroblastic foci, honey-            low-up CT. The CT findings were in-                        subpleural sparing), presence of fibro-
comb change, and normal lung coexist-            terpreted on the basis of the recom-                      sis in upper lobes, and basal and pe-
ing in variable proportions (1). The di-         mendations of the nomenclature com-                       ripheral predominance of fibrosis
agnosis of NSIP was made on the basis            mittee of the Fleischner Society (20).                    were also assessed. Subpleural refers
of the presence of temporally uniform                Simple dichotomous presence vari-                     to the region immediately adjacent to
interstitial inflammation and/or fibrosis,         ables were assessed at CT. GGO was                        the costal pleura (ie, located 1 cm
with subdivision into cellular (n      4),       defined as an area of hazy increased                       from the pleura), whereas peripheral
mixed (n        10), and fibrotic (n     9)       attenuation without obscuration of un-                    refers to the outer one-third of the lung.
types according to current diagnostic            derlying vascular markings. Intralobular                      Observations made by using semi-
criteria (1).                                    reticular opacity was considered to be                    quantitative scores include the extents
                                                 present when interlacing line shadows                     of GGO, consolidation, reticulation, and
CT Scanning Protocol                             were seen in secondary pulmonary lob-                     honeycombing, each scored as grade 0,
Several CT scanners were used for this           ules. Traction bronchiectasis and bron-                   absent; grade 1, involving 1%– 4% of
study: Two scanners were from one                chiolectasis were defined as irregular                     the lung parenchyma; grade 2, involving
manufacturer (HiSpeed Advantage and              bronchial and bronchiolar dilatation in ar-               5%–25% of the lung parenchyma; grade
LightSpeed 16; GE Healthcare, Milwau-            eas with parenchymal abnormality. Hon-                    3, involving 26%–50% of the lung pa-
kee, Wis), three multidetector scanners          eycombing was considered to be present                    renchyma; or grade 4, involving more
were from another manufacturer (So-              when clustered cystic air spaces with                     than 50% of the lung parenchyma. Hon-
matom [Volume Zoom] 4-section and                well-defined and thick walls were seen.                    eycombing cysts were categorized as
Sensation 16- and 64-section; Siemens                The anatomic distribution was                         grade 0, absent; grade 1, 1–5 mm; or
Medical Solutions, Erlangen, Ger-                classified as peribronchovascular (a                       grade 2, larger than 5 mm.
many), and one electron beam scanner
was from a third manufacturer (C-150L;
                                                   Table 1
Imatron, San Francisco, Calif). The im-
ages were obtained by using 1- or 2-mm              Criteria for Diagnosis of NSIP and IPF at Thin-Section CT
collimation at 10-mm intervals or volu-
metrically on multidetector CT scanners
                                                    Diagnosis   Definite                                            Probable
with 0.6- or 1-mm collimation and
1-mm reconstruction. The scans were                  NSIP       Extensive GGO, no or only mild reticulation,      Moderate GGO with overall extent greater than
obtained with the patient in supine po-                            traction bronchiectasis, no or only               the extent of reticulation, traction
sition at full inspiration and were recon-                         minimal honeycombing, basal                       bronchiectasis, no or only minimal
structed by using a high-spatial-fre-                              predominance of abnormalities, relative           honeycombing, basal predominance of
quency algorithm. All images were                                  subpleural sparing                                abnormalities
viewed at window settings optimized for              IPF        Reticulation in all lobes, moderate or            Bilateral reticulation, mild honeycombing, mild
assessment of lung parenchyma (width,                              extensive honeycombing, no or minimal             to moderate GGO with overall extent of GGO
1000 –1500 HU; level, 600 to 700                                   GGO, peripheral and basal                         being less than the extent of reticulation,
                                                                   predominance of abnormalities                     peripheral and basal predominance of
HU). Hardcopy images were available
for 19 patients. In the remaining 29 pa-

Radiology: Volume 247: Number 1—April 2008                                                                                                                          253
THORACIC IMAGING: Nonspecific Interstitial Pneumonia: Changes over Time                                                                                                     Silva et al

    The size of honeycomb cysts was                          tion). The observations of the two ra-                         diagnosis as being definite (high) or
determined given the largest visual-                         diologists were summed for the pur-                            probable (low), according to specific
ized cyst. The predominant CT pat-                           poses of analysis; a nine-point scale                          diagnostic criteria (Table 1). When
tern, which was classified as grade 0,                        (score of 0 – 8) was used for extents of                       the CT findings did not match any of
predominantly inflammatory (if the                            GGO, consolidation, reticulation, and                          the criteria, the first-choice diagnosis
extent of ground-glass attenuation                           honeycombing and a five-point scale                             was considered as indeterminate. Dis-
and/or consolidation was greater than                        (score of 0 – 4) was used for the size of                      agreements on presence variables and
the extent of reticulation and/or hon-                       honeycomb cysts and the predomi-                               on the likely thin-section CT diagnosis
eycombing); grade 1; equivalent fi-                           nance of the reticular pattern.                                were resolved by consensus review of
brotic and inflammatory; or grade 2,                              Following the initial assessment of                        divergent scores.
predominantly fibrotic (if the extent of                      the thin-section CT images, each radi-
reticulation and/or honeycombing was                         ologist made a first-choice diagnosis of                        Statistical Analysis
greater than the extent of ground-                           NSIP or IPF for each patient and                               Results are given as means and stan-
glass attenuation and/or consolida-                          graded the degree of confidence in this                         dard deviations or medians and ranges

 Figure 1

 Figure 1: Transverse thin-section lung CT scans of characteristic NSIP findings in 40-year-old woman. Initial scans at levels of (a) bronchus intermedius and
 (b) lower zones show bilateral GGO and minimal reticulation. Note relative subpleural sparing (arrowheads) immediately adjacent to pleura in dorsal regions of lower lobes and
 traction bronchiectasis and bronchiolectasis. (c, d) Follow-up scans 45 months later show increase in extent of reticulation and traction bronchiectasis and bronchiolectasis. Min-
 imal subpleural honeycombing has developed (arrows) and relative subpleural sparing is still apparent (arrowheads). All scans were interpreted as being consistent with NSIP.

254                                                                                                                                      Radiology: Volume 247: Number 1—April 2008
THORACIC IMAGING: Nonspecific Interstitial Pneumonia: Changes over Time                                                                                         Silva et al

  Figure 2

  Figure 2: Transverse thin-section lung CT scans show characteristic UIP findings in 53-year-old woman. Initial scans at levels of (a) right upper lobe bronchus and
  (b) lower zones show reticular pattern and traction bronchiectasis involving mainly peripheral lung regions and bases. Note mild honeycombing (arrowheads) and mini-
  mal GGO. (c, d) Follow-up scans 67 months later show progression of reticulation and honeycombing in predominantly peripheral and basal distribution. All scans were
  interpreted as being consistent with UIP.

(for nonnormally distributed variables).                 0.21– 0.40; moderate,         0.41– 0.60;                 were examined by using the Spearman
Interobserver variation in CT findings                    good,      0.61– 0.80; and excellent,                     rank correlation coefficient ( ). The Wil-
was quantified by using the unweighted                    0.81–1.00 (21).                                           coxon rank sum test was used for un-
kappa coefficient ( ) of agreement for                         Comparisons between baseline and                     paired group comparisons of semiquan-
presence variables and the weighted                      follow-up thin-section CT findings posi-                   titative scales and the Wilcoxon signed
kappa coefficient ( w) of agreement                       tive for NSIP and IPF were made by                        rank test for paired comparisons of
(quadratic weighting) for variables scored               using 2 statistics or a two-tailed Fisher                 data. The predictive value of baseline
by using semiquantitative scales. The in-                exact test (when the smallest of the four                 CT features in NSIP patients against
terobserver agreement was classified as                   expected numbers was less than five).                      evolution to an IPF-like appearance
follows: poor,        0 – 0.20; fair,                    Interrelationships between CT variables                   were assessed by means of univariate

Radiology: Volume 247: Number 1—April 2008                                                                                                                           255
THORACIC IMAGING: Nonspecific Interstitial Pneumonia: Changes over Time                                                                                                       Silva et al

logistic regression. A P value of less                              Given the lack of major disagreement                             8%; and three of 25, 12%, respectively)
than .05 was considered to indicate a                               between the two radiologists, analyses                           (Figs 1, 2; Table 2). NSIP was also char-
significant difference.                                              of the extents of individual patterns                            acterized by greater extent of GGO (P
                                                                    were made by using summed scores.                                .002), lower extent of honeycombing
                                                                                                                                     (P     .02), smaller size of honeycomb
 Results                                                            Comparison between NSIP and IPF                                  cysts (P .01), and a lower percentage
                                                                    Findings at Initial CT                                           of patients with predominant reticular
Thin-Section CT Findings                                            NSIP patients were characterized by a                            pattern (P     .01) (Table 3). Variables
There was moderate to excellent inter-                              lower prevalence of honeycombing (five                            that were almost invariably positive
observer agreement in the interpreta-                               of 23, 22%) (P         .01) and peripheral                       (presence of reticulation, presence of
tion of simple dichotomous presence                                 predominance of fibrosis (16 of 23,                               GGO, traction bronchiectasis and bron-
variables (     0.56 –1.00) and good to                             70%) (P       .01), higher prevalence of                         chiolectasis, upper lobe fibrosis, basal
excellent agreement for continuous semi-                            relative subpleural sparing (10 of 23,                           predominance of fibrosis, and lower
categorical variables ( w         0.73–                             43%) (P      .005), and a random distri-                         lobe predominance of disease) or nega-
1.00). There was excellent interob-                                 bution of disease (13 of 23, 57%) (P                             tive (presence of consolidation) were
server agreement ( w       0.90) on the                             .005) as compared with IPF patients (14                          not evaluated further.
likely CT diagnosis of NSIP or IPF.                                 of 25, 56%; 24 of 25, 96%; two of 25,
                                                                                                                                     Interrelationships between Selected
 Table 2                                                                                                                             Thin-Section CT Patterns at Baseline
                                                                                                                                     In the combined patient population
   Baseline Thin-Section CT Findings of Patients with NSIP and IPF                                                                   and separately in the IPF and NSIP
   CT Finding                                                         NSIP (n     23)          IPF (n     25)          P Value       patient subsets, the extent of reticula-
                                                                                                                                     tion was not linked to either the extent
      Reticulation                                                    21 (91)                  24 (96)                 NS
                                                                                                                                     of GGO or of honeycombing. By con-
      GGO                                                             23 (100)                 24 (96)                 NS
                                                                                                                                     trast, there was a strong negative rela-
      Consolidation                                                    5 (22)                   1 (4)                    .09*
                                                                                                                                     tionship (         0.69, P     .001) be-
      Honeycombing                                                     5 (22)                  14 (56)                 .01
      Traction bronchiectasis                                         21 (91)                  23 (92)                 NS
                                                                                                                                     tween the extent of GGO and that of
      Traction bronchiolectasis                                       21 (91)                  23 (92)                 NS            honeycombing, which remained signif-
      Relative subpleural sparing                                     10 (43)                   2 (8)                    .005        icant in both diseases when analyzed
      Lower zone predominance of abnormalities                        19 (83)                  22 (88)                 NS†           separately (IPF,         0.71, P .001;
      Presence of upper lobe fibrosis                                  21 (91)                  22 (88)                 NS            NSIP,         0.54, P .01).
      Peripheral predominance of fibrosis                              16 (70)                  24 (96)                   .01
      Basal predominance of fibrosis                                   19 (83)                  21 (84)                 NS            Comparison between NSIP and IPF
      Anatomic distribution                                                                                                          Findings at Follow-up CT
         Regional (peripheral and peribronchovascular)                10 (43)                  22 (88)                    .005       Variables that differed significantly be-
         Random                                                       13 (57)                   3 (12)                    .005       tween the two diseases at baseline were
                                                                                                                                     compared at follow-up thin-section CT.
   Note.—Unless otherwise indicated, data are the number of patients. Data in parentheses are percentages. NS, not significant.
   P .05 were significant.                                                                                                            Group differences that remained signif-
   * Calculated with Fisher exact test.                                                                                              icant were the higher prevalence of rel-
     Distribution was random in the remaining patients; none of the patients had an upper lobe predominance of disease. The          ative subpleural sparing in NSIP (nine of
   distribution of disease was peribronchovascular in one IPF patient and one NSIP patient. The distribution was peripheral in the   23, 39%) than in IPF (0%) (P       .001)
   remaining patients with a regional demarcation of disease (NSIP, n 9; IPF, n 21).
                                                                                                                                     patients, the greater prevalence (23 of
                                                                                                                                     25, 92%) (P .03) and extent (P .01)
                                                                                                                                     of honeycombing in IPF than in NSIP
 Table 3                                                                                                                             patients, and the larger size of honey-
                                                                                                                                     comb cysts in IPF patients (P       .02).
   Extent of CT Findings at Baseline Study                                                                                           The extent of honeycombing increased
   CT Score                                               NSIP                           UIP                           P Value       between the initial and follow-up scans
                                                                                                                                     in NSIP (P     .005) and IPF (P    .001)
      Extent of honeycombing                              0 (0–2)                        1 (0–6)                       .02
                                                                                                                                     groups (Figs 1, 2). In contrast, baseline
      Size of honeycombing cysts                          0 (0–4)                        1 (0–4)                       .01
      Reticular predominance                              1 (0–4)                        4 (0–4)                       .01
                                                                                                                                     differences in the extent of GGO were
      Extent of reticulation                              4 (0–6)                        4 (0–6)                       NS            absent at follow-up, reflecting a marked
      Extent of GGO                                       6 (2–8)                        3 (1–8)                       .002          decrease in the extent of GGO in NSIP
                                                                                                                                     between the baseline and follow-up
   Note.—Data reflect median values of summed scores with ranges in parentheses. The extent of consolidation was not                  scans (P     .001), as compared with a
   evaluated owing to the low prevalence of consolidation. NS, not significant. P .05 indicates significant difference.
                                                                                                                                     minor reduction in the extent of GGO in

256                                                                                                                                          Radiology: Volume 247: Number 1—April 2008
THORACIC IMAGING: Nonspecific Interstitial Pneumonia: Changes over Time                                                                                  Silva et al

                                                   Figure 3
IPF (P .05) (Fig 3). The prevalence of
a random distribution of disease in NSIP
was also absent at follow-up; as seen in
a subset of NSIP patients, disease was
random in distribution at baseline but
became more peripheral at follow-up.
Similarly, baseline differences in scores
for the predominance of a reticular pat-
tern were absent at follow-up.
     At follow-up, there was no differ-
ence between the predominant CT pat-
tern scores for IPF (median, 4; range,
0 – 4) and NSIP (median, 4; range, 0 – 4)
(P     .67). Across the 48 patients, the
grade did not change in 19 (NSIP, n 8;
IPF, n 11), it became relatively more
inflammatory in three (all IPF), but
became relatively more fibrotic in 26
(NSIP, n 15; IPF, n 11). Change in
pattern score differed between NSIP
(median, 2; range, 0 – 4) and IPF (me-
dian, 0; range, 2 to 4) (P       .05). On
paired testing, the pattern score at fol-
low-up (median, 4; range, 0 – 4) was               Figure 3: Transverse thin-section lung CT of characteristic NSIP findings at level of inferior pulmonary
strikingly more fibrotic than at initial CT         vein of right lung in 41-year-old man. (a) Initial scan shows extensive GGO and minimal reticulation. Note
(median, 2; range, 0 – 4) (P .001). For            relative subpleural sparing (arrowheads) and mild traction bronchiectasis. (b) Follow-up scan 67 months later
NSIP, this trend was strikingly signifi-            shows decrease in extent of GGO and increase in extent of reticulation. Relative subpleural sparing persists
cant (P .001). For IPF, this trend was             (arrowheads) and traction bronchiectasis is more evident. All scans were interpreted as being consistent with
weak (P .03).                                      NSIP.

Diagnostic Thin-Section CT Features at
Baseline and at Follow-up                        line CT and 12 at follow-up CT. Five                      months). The most striking finding at
At presentation, thin-section CT find-            (28%) of 18 patients with initial CT                      follow-up was that the differences be-
ings seen in IPF patients were inter-            findings suggestive of NSIP developed                      tween the two diseases do diminish be-
preted as suggestive of IPF in 11 (44%)          findings interpreted as suggestive of                      cause the findings in a proportion of
of 25 patients (definite IPF, n 6; prob-          IPF (Fig 4) and one as indeterminate                      patients with NSIP come to resemble
able IPF, n     5), indeterminate in five         at follow-up CT. On univariate logistic                   those of IPF. At follow-up CT, there is a
(20%) patients, and suggestive of NSIP           regression, no individual thin-section                    decrease in the extent of GGO, increase
in nine (36%) patients (definite NSIP,            CT feature at baseline was found to                       in the extent of reticulation and honey-
n 8; probable NSIP, n 1) by consen-              distinguish between these two sub-                        combing in NSIP, development of cysts,
sus reading by the two thoracic radiolo-         groups. The patients that progressed from                 and increase in honeycombing and re-
gists. Thin-section CT findings seen in           an NSIP pattern to a UIP pattern had                      ticulation. Although the extent of retic-
NSIP patients were suggestive of NSIP            mixed (n 2) or fibrotic (n 3) NSIP at                      ulation and honeycombing of NSIP re-
in 18 (78%) of 23 (definite NSIP, n               lung biopsy; none of these patients had                   main less severe than those of IPF, both
16; probable NSIP, n       2), indetermi-        cellular NSIP.                                            diseases increase fibrotic elements in
nate in three (13%), and suggestive of                The intervals between scans did                      the same direction, but this has a more
IPF in two (9%) (definite IPF, n        1;        not vary significantly with changes in                     radical influence on NSIP scores as
probable IPF, n 1) patients.                     CT findings, either across the whole                       these were less IPF-like from the start.
    Of 23 NSIP patients, two were in-            cohort or in either disease subset                             Overall, among patients with NSIP,
terpreted as suggestive of IPF and               (all P    .1).                                            18 had initial abnormalities suggestive
three as indeterminate at baseline CT;                                                                     of NSIP at thin-section CT. At follow-up,
six patients were interpreted as sug-                                                                      the pattern remained as NSIP in 12 of
gestive of IPF and five as indetermi-               Discussion                                              these 18 patients but evolved to an inde-
nate at follow-up CT. Overall, 18 of             To our knowledge, our study is the first                   terminate pattern in one patient and a
these patient findings were inter-                to compare the CT findings of NSIP and                     pattern resembling IPF in the remaining
preted as suggestive of NSIP at base-            IPF at long-term follow-up (34 –155                       five. No individual thin-section CT at

Radiology: Volume 247: Number 1—April 2008                                                                                                                    257
THORACIC IMAGING: Nonspecific Interstitial Pneumonia: Changes over Time                                                                                          Silva et al

baseline was found to distinguish be-                    minority of NSIP patients exhibited                        of ground-glass attenuation rather than
tween these two subgroups.                               early deterioration despite treatment,                     reticulation, absent or relatively mild
    The results of our study suggest that                and this was predictive of an IPF-like                     honeycombing, and greater likelihood
there are two separate NSIP subgroups,                   course to a fatal outcome (5). However,                    of random distribution of findings as
with one subgroup representing NSIP                      in our study there was no appreciable                      compared with IPF. These results are
progressing to a pattern that resembles                  difference in the pattern, extent, or dis-                 similar to those of previous studies (10–
IPF and one subgroup that maintains                      tribution of findings between patients                      12,14). On the basis of anecdotal expe-
the overall features of NSIP even with                   with NSIP that maintained an NSIP pat-                     rience, it was recently suggested that
progression of fibrosis. This observation                 tern at follow-up and those that changed                   relative subpleural sparing may help
is concordant with the findings of Latsi                  to a pattern resembling IPF.                               distinguish NSIP from IPF (2). This was
et al (5) in a study of serial pulmonary                     Our study shows that at initial pre-                   confirmed in our study, which showed
function change against mortality in IPF                 sentation NSIP is characterized by                         relative sparing of the lung immediately
and idiopathic NSIP. In that study, a                    greater extent of GGO, predominance                        adjacent to the pleura in the dorsal lung

 Figure 4

 Figure 4: Transverse thin-section lung CT in 61-year-old man with biopsy-proved NSIP. Initial scans at levels of (a) bronchus intermedius and (b) lower zones show
 GGO in lower lobe predominance, minimal traction bronchiectasis, and mild reticulation and were interpreted as being characteristic of NSIP. (c, d) Follow-up scans 83
 months later show increase in extension of reticulation, worsening of the traction bronchiectasis, and development of honeycombing. Note reduction in extent of GGO.
 Follow-up scan was interpreted as suggestive of IPF.

258                                                                                                                             Radiology: Volume 247: Number 1—April 2008
THORACIC IMAGING: Nonspecific Interstitial Pneumonia: Changes over Time                                                                                  Silva et al

regions of the lower lobes in 10 (43%)           tation that allow distinction between pa-                   10. MacDonald SL, Rubens MB, Hansell DM,
of 23 patients with NSIP compared with           tients with NSIP that maintain an NSIP                          et al. Nonspecific interstitial pneumonia and
                                                                                                                 usual interstitial pneumonia: comparative
two (8%) of 25 patients with IPF.                pattern at follow-up and those that
                                                                                                                 appearances at and diagnostic accuracy of
    Our study had limitations. It was            progress to an IPF pattern; in other                            thin-section CT. Radiology 2001;221(3):
retrospective, included a small number           words, similar initial CT findings positive                      600 – 605.
of patients who underwent CT with dif-           for NSIP may have different outcomes.
                                                                                                             11. Hartman TE, Swensen SJ, Hansell DM, et al.
ferent scanners, and had a small inter-                                                                          Nonspecific interstitial pneumonia: variable
val between the initial and follow-up CT         References                                                      appearance at high-resolution chest CT.
examinations. Because lung biopsy is              1. American Thoracic Society; European Respi-                  Radiology 2000;217(3):701–705. [Published
seldom obtained from patients with                   ratory Society. American Thoracic Society/                  correction appears in Radiology 2001;
characteristic thin-section CT findings               European Respiratory Society International                  218(2):606.]
positive for IPF, the study is biased to-            Multidisciplinary Consensus Classification of            12. Screaton NJ, Hiorns MP, Lee KS, et al. Se-
                                                     the Idiopathic Interstitial Pneumonias. This
ward patients with atypical IPF findings.                                                                         rial high resolution CT in non-specific inter-
                                                     joint statement of the American Thoracic So-                stitial pneumonia: prognostic value of the ini-
The study did not compare the CT with                ciety (ATS), and the European Respiratory                   tial pattern. Clin Radiol 2005;60(1):96 –104.
the clinical or functional findings at diag-          Society (ERS) was adopted by the ATS board
nosis or follow-up and did not assess the            of directors, June 2001 and by the ERS Exec-            13. Kim TS, Lee KS, Chung MP, et al. Nonspecific
effect of therapy. The study included a              utive Committee, June 2001. Am J Respir                     interstitial pneumonia with fibrosis: high-resolu-
small number of patients with NSIP who               Crit Care Med 2002;165(2):277–304. [Pub-                    tion CT and pathologic findings. AJR Am J
                                                     lished correction appears in Am J Respir Crit               Roentgenol 1998;171(6):1645–1650.
had follow-up CT. Although five (28%)
                                                     Care Med 2002;166(3):426.]
of 18 patients with initial findings sug-                                                                     14. Elliot TL, Lynch DA, Newell JD Jr, et al.
gestive of NSIP had follow-up CT scans            2. Kim DS, Collard HR, King TE Jr. Classifica-                  High-resolution computed tomography fea-
                                                     tion and natural history of the idiopathic in-              tures of nonspecific interstitial pneumonia
that were more suggestive of IPF, this
                                                     terstitial pneumonias. Proc Am Thorac Soc                   and usual interstitial pneumonia. J Comput
may not be representative of the entire              2006;3(4):285–292.                                          Assist Tomogr 2005;29(3):339 –345.
NSIP population owing to potential bias
                                                  3. Nicholson AG, Colby TV, du Bois RM,                     15. Flaherty KR, Mumford JA, Murray S, et al.
in patient selection.
                                                     Hansell DM, Wells AU. The prognostic sig-                   Prognostic implications of physiologic and
    Furthermore, although initial thin-                                                                          radiographic changes in idiopathic intersti-
                                                     nificance of the histologic pattern of inter-
section CT findings characteristic of                 stitial pneumonia in patients presenting                    tial pneumonia. Am J Respir Crit Care Med
NSIP progressed to a pattern of UIP,                 with the clinical entity of cryptogenic fibro-               2003;168(5):543–548.
this change in pattern does not neces-               sing alveolitis. Am J Respir Crit Care Med
                                                                                                             16. Kim EY, Lee KS, Chung MP, Kwon OJ, Kim
sarily reflect a change in histologic pat-            2000;162(6):2213–2217.
                                                                                                                 TS, Hwang JH. Nonspecific interstitial pneu-
tern of NSIP to UIP. All five patients             4. Travis WD, Matsui K, Moss J, Ferrans VJ. Idio-              monia with fibrosis: serial high-resolution
who progressed from an NSIP to a UIP                 pathic nonspecific interstitial pneumonia: prog-             CT findings with functional correlation. AJR
pattern had mixed or fibrotic NSIP at                 nostic significance of cellular and fibrosing pat-            Am J Roentgenol 1999;173(4):949 –953.
biopsy; none of the four NSIP patients               terns. Am J Surg Pathol 2000;24(1):19–33.
                                                                                                             17. Nishiyama O, Kondoh Y, Taniguchi H, et al.
with a cellular pattern at biopsy pro-            5. Latsi PI, du Bois RM, Nicholson AG, et al.                  Serial high resolution CT findings in nonspe-
gressed to an IPF pattern at CT. Pro-                Fibrotic idiopathic interstitial pneumonia:                 cific interstitial pneumonia/fibrosis. J Com-
                                                     the prognostic value of longitudinal func-                  put Assist Tomogr 2000;24(1):41– 46.
spective studies in a larger number of
                                                     tional trends. Am J Respir Crit Care Med
patients with close correlation between                                                                      18. Akira M, Inoue G, Yamamoto S, Sakatani M.
the CT, clinical, and histologic findings                                                                         Non-specific interstitial pneumonia: findings
                                                  6. Nicholson AG, Wells AU. Nonspecific inter-                   on sequential CT scans of nine patients. Tho-
will be required to further assess the
                                                     stitial pneumonia: nobody said it’s perfect.                rax 2000;55(10):854 – 859.
possibility that in some patients NSIP
                                                     Am J Respir Crit Care Med 2001;164(9):                  19. Jeong YJ, Lee KS, Muller NL, et al. Usual
may progress to UIP.                                 1553–1554.
    In summary, at presentation, NSIP                                                                            interstitial pneumonia and non-specific inter-
                                                  7. Katzenstein AL, Zisman DA, Litzky LA,                       stitial pneumonia: serial thin-section CT
is characterized by the predominance of                                                                          findings correlated with pulmonary function.
                                                     Nguyen BT, Kotloff RM. Usual interstitial
GGO, lack of honeycombing, presence                                                                              Korean J Radiol 2005;6(3):143–152.
                                                     pneumonia: histologic study of biopsy and
of relative subpleural sparing, and low              explant specimens. Am J Surg Pathol 2002;                                  ¨
                                                                                                             20. Austin JH, Muller NL, Friedman PJ, et al.
prevalence of peripheral distribution of             26(12):1567–1577.                                           Glossary of terms for CT of the lungs: recom-
disease, findings which allow distinction                                                                         mendations of the Nomenclature Committee
                                                  8. Lynch DA, Travis WD, Muller NL, et al.
from IPF in the majority of patients. At             Idiopathic interstitial pneumonias: CT fea-                 of the Fleischner Society. Radiology 1996;
3 years or longer follow-up, 28% of pa-              tures. Radiology 2005;236(1):10 –21.                        200(2):327–331.
tients with initial CT findings suggestive                                                                    21. Brennan P, Silman A. Statistical methods for
                                                  9. Hunninghake GW, Zimmerman MB, Schwartz
of NSIP progress to findings suggestive               DA, et al. Utility of a lung biopsy for the diagnosis       assessing observer variability in clinical mea-
of IPF. Given the results of our study,              of idiopathic pulmonary fibrosis. Am J Respir                sures. BMJ 1992;304(6840):1491–1494.
there are no CT features seen at presen-             Crit Care Med 2001;164(2):193–196.

Radiology: Volume 247: Number 1—April 2008                                                                                                                    259

To top