Upload
jm-anderson
View
217
Download
2
Embed Size (px)
Citation preview
CT-guided Lung Biopsy: Factors InfluencingDiagnostic Yield and Complication Rate
J . M. ANDERSON, J. MURCHISON, D. PATEL
Department of Radiology, The Royal Infirmary of Edinburgh, Edinburgh, UK
Received: 6 February 2003 Revised: 17 April 2003 Accepted: 26 April 2003
AIM: To determine factors influencing diagnostic yield in computed tomography (CT)-guided biopsyof lung lesions.MATERIALS AND METHODS: One hundred and ninety-five consecutive CT-guided lung biopsieswere performed in 182 patients between August 1995 and September 2000 and either fine-needleaspirate samples for cytology or core biopsy samples for histology were collected. Procedures weredivided into a diagnostic group (true-positive and true-negative results) and a non-diagnostic group(false-positive and false-negative results) and the factors affecting diagnostic accuracy assessed.RESULTS: One hundred and fifty-six lesions (86%) were malignant, and 26 (14%) were benign.More than one biopsy was performed for 12 lesions. One hundred and thirty-two biopsies were true-positive, 27 true-negative and 36 false-negative. No false-positive results occurred in the study.Overall diagnostic accuracy was 81.5%. Significantly more core biopsies than fine-needle aspirateswere diagnostic: 93 versus 78% ðp < 0:005Þ: No difference was found in frequency of pneumothoraxbetween these two groups. There was a difference in the average depth from the pleural surface oflesions in the diagnostic and non-diagnostic groups, but this did not attain statistical significance: 9.8versus 17.2 mm ðp 5 0:054Þ:CONCLUSION: In this study CT-guided lung biopsy core biopsy was a more accurate method oftissue sampling than fine-needle aspiration, and was not associated with an excess of complications.Anderson, J. M. et al. (2003). Clinical Radiology 58: 791–797.
q 2003 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Key words: lung, biopsy, computed tomography, guidance, accuracy, pneumothorax.
INTRODUCTION
Computed tomography (CT)-guided lung biopsy was firstdescribed in 1976 [1]. Since then, after a great deal of work torefine its safety and diagnostic accuracy, it has become a widelyused diagnostic tool in the management of patients withsuspected lung cancer. Diagnostic accuracies of 82–96% havebeen reported [2,3] with acceptable complication rates; theincidence of pneumothorax is reported as 22–45% [3–5].Traditionally, fine-needle aspirate biopsies were performed,because of the unacceptable morbidity and mortality associatedwith pulmonary haemorrhage after core biopsy [6,7]. Morerecently, the availability of smaller gauge, spring-loaded corebiopsy needles has revived interest in their use, given thereduced incidence of pulmonary haemorrhage and the acceptedlimitations of cytological analysis of fine-needle aspiration(FNA) samples. One main limitation of aspiration biopsies is
the operator’s inability to assess the adequacy of the samplevisually. The presence of a cytopathologist, to confirm thatdiagnostic material has been obtained, has been shown toimprove diagnostic yield [8], but if this resource is notavailable, it has been suggested that core biopsies should beobtained routinely [9]. Studies to establish other factors thataffect diagnostic yield have been limited, although there hasbeen much more published on the factors affecting theincidence of pneumothorax. We undertook this study of 5years of our own experience to assess factors affectingdiagnostic accuracy, with specific reference to the differencesbetween core biopsy (which was used increasingly during thisperiod) and FNA biopsy. Factors affecting the frequency ofpneumothorax were also studied.
MATERIALS AND METHODS
Between August 1995 and September 2000, 195 CT-guidedlung biopsies were performed on 182 patients, 11 patientshaving two biopsies and one patient having three biopsies.
0009-9260/03/$30.00/0 q 2003 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Guarantor and correspondent: J. M. Anderson, Department of ClinicalRadiology, New Royal Infirmary of Edinburgh, Little France, EdinburghEH 3 9YW, UK. Tel: þ44-131-536-2900; fax: þ44-131-242-3702; E-mail:[email protected]
Clinical Radiology (2003) 58: 791–797doi:10.1016/S0009-9260(03)00221-6, available online at www.sciencedirect.com
These cases were identified retrospectively using the RadiologyInformation System (Ricketts System).
Patients were referred by the respiratory physicians aftermultidisciplinary discussion at weekly clinical meetings.Patients with severe chronic obstructive pulmonary disease(COPD) (FEV1 ,1 l) or previous pneumonectomy, thoseunable to lie flat or suspend respiration and the very infirm wereexcluded. Clotting studies were performed before the pro-cedure. Informed consent was obtained by the respiratoryphysicians and reiterated by the radiologist in charge of theprocedure. All procedures were performed by one of threeconsultants experienced in the technique, or by a registrar intraining under their direct supervision. All patients underwentstaging CT of the chest and upper abdomen before arriving forthe biopsy procedure. The patient was positioned in the CTscanner (GE Medical Systems, HiSpeed1.X) as appropriate,given the trajectory chosen from the staging images. Instructionwas given in breath-holding technique. An initial scan wastaken through the region of interest using a slice thickness of5 mm. These images were used by the operator to decide on asuitable approach to the lesion. Localization was performedusing CT images with laser lighting and skin markers. Usingaseptic technique, local anaesthetic (lignocaine 1%) wasinfiltrated to the pleura. The operator then decided whichneedle to use for the procedure, selecting either a spinal needleof 20 or 22 gauge (Becton Dickinson Spinal Needle, Quinketype point), to obtain a fine-needle aspirate, or an 18 or 20gauge core biopsy needle (Gallini Medical Devices biopsyneedle) to obtain a core of tissue. The needle was thenintroduced through the skin and subcutaneous tissues. Itsposition was checked on CT before it was introduced to themeasured depth of the lesion. Once the tip had been shown tolie within the pathological lesion, the first tissue sample wasobtained. The operator then assessed the apparent adequacy ofthe sample before deciding to proceed with further passes. Thiswas assessed by visual inspection, as facilities for immediatecytological assessment are not available in our institution.Further passes were performed as required. Once the operatorwas satisfied with the samples obtained, immediate post-biopsyCT was obtained to check for pneumothorax. Assuming thepatient was clinically stable they were returned to theInvestigations Unit for post-procedure monitoring. All patientshad a chest radiograph performed at 4 h, or sooner if theybecame symptomatic. After this, each patient was reviewed byone of the respiratory physicians before being discharged on thesame evening if no complications had ensued. Cytopathologysamples were subsequently inspected by one of a number ofpathologists and results issued to the respiratory staff managingthe patient between 3 and 5 days later.
Data Collection
Data were collected in five categories: patient demo-graphics, details of the nature of the lesion, technical detailsregarding the procedure itself, details of complications thatensued and their treatment, and details regarding diagnosticyield. The patients’ age and sex were recorded. Lesion size,location and depth from the pleural surface were noted. Thesize of the lesion was measured as the maximum diameter (in
millimetres) on the hard copy of CT images obtained in 5 mmslices immediately pre-biopsy printed on mediastinal windowsettings. Depth of the lesion was measured from the pleuralsurface at the point of needle puncture, along the needle track,to the tip of the lesion. Procedural details, taken from the formalreports in the computer database, included the type and size ofneedle used, the number of passes made, and the status of theoperator, i.e. consultant or registrar in training. Complicationsrecorded were the presence of an immediate pneumothorax onthe post-biopsy CT images, the presence of a pneumothorax onthe chest radiograph at 4 h, whether any treatment was requiredif pneumothorax was present, and any other complications thatensued. Details of treatment required for post-proceduralpneumothoraces and other complications were obtained fromreview of the case notes. Diagnostic data included the result ofthe initial biopsy, details of other diagnostic proceduresperformed, and the result of additional procedures, includingpneumonectomy or autopsy where relevant. These details wereobtained directly from the computer database in the pathologydepartment. Patients whose biopsies revealed pleural-basedtumours, such as mesothelioma or pleural-based metastases,were excluded from the study, as it was thought that the moreadvanced nature of these diseases would introduce bias.
RESULTS
One hundred and ninety-five biopsies were performed on182 patients: 11 patients had two procedures and one patienthad three procedures. One hundred and three patients weremale and 79 were female. The average age was 67.5 years(range 29–87 years). The average size of the lesions was41 mm (range 8–100 mm). One hundred and two lesions werecontiguous with the pleural surface: the average depth of allother lesions from the pleural surface was 24 mm (range 5–70 mm). One hundred and fifty-one FNA biopsies and 44 corebiopsies were performed. On average, 1.8 passes were madeacross the pleural surface (range 1–4). One hundred and fifty-six cases (86%) had a final diagnosis of malignancy and 26cases (14%) had a final benign diagnosis. Specific diagnoses arelisted in Table 1.
Eleven patients underwent two CT-guided lung biopsies andone patient underwent three separate procedures. All had repeatprocedures because the clinical suspicion of malignancy washigh, and no malignant cells were initially returned. In fourpatients, including the patient who had three procedures, theultimate diagnosis was of a benign process, as evidenced by therepeat biopsies and the subsequent clinical course. Eightpatients were given a final diagnosis of malignancy: seven ofthese were diagnosed accurately by the second CT-guided lungbiopsy. One patient had a further inconclusive biopsyperformed, and proceeded to lobectomy, which showedadenocarcinoma.
To determine factors affecting diagnostic yield, the caseswere divided into two groups. The “successful” group ofdiagnostic biopsies comprised the true-positive and true-negative biopsies. These were defined as cases where theinitial result of the CT-guided biopsy (broadly speaking, benignor malignant) was the same as the final diagnosis. A finaldiagnosis of malignancy was recorded when supported by
CLINICAL RADIOLOGY792
histology results from pneumonectomy or autopsy, or when the
patients’ clinical course was typical of malignant disease, with
either the progression of the primary intrapulmonary lesion or
development of metastatic lesions. A final diagnosis of benign
disease was recorded when either subsequent histological
samples concurred or when the patients’ clinical course was
typical of a benign process with no progression or resolution of
the primary intrapulmonary lesion on subsequent CT. The
“failure” group of non-diagnostic biopsies comprised the false-
negative biopsies, where initial results showed either insuffi-
cient samples or only benign cells, and the final diagnosis was
shown to be malignant either by repeat biopsy or by a typical
clinical course. Using these criteria, 132 biopsies (68%) were
true-positive, 27 (14%) were true-negative, and 36 (18%) were
false-negative. There were no false-positive biopsies in the
study. This gives an overall accuracy of 81.5%, with sensitivity
of 84.6% and specificity of 100%.
Data were then analysed to assess which factors affected
diagnostic yield. Five factors were investigated, namely lesion
size, lesion depth from the pleural surface, type of biopsy
needle, number of passes made through the pleural surface and
the status of the operator. Results are presented in Table 2.
The data regarding lesion size, depth and number of passes
made were analysed using Student’s unpaired t-test assuming
unequal variance. The data regarding the type of biopsy needle
and the status of the operator were analysed using the chi-
squared test. A p-value of less than 0.05 was considered
significant.
A significant increase in accuracy was noted with core
biopsies when compared with fine needle aspirate biopsies,
despite the smaller number of patients in this group. A trend
towards increased accuracy was noted with proximity of thelesion to the pleural surface, but this did not reach statistical
significance. Otherwise, no significant differences were
revealed.
The main complication of CT-guided lung biopsy is the
development of a pneumothorax. We documented the presence
of an immediate pneumothorax on the post-biopsy CT images,
and the presence of a delayed pneumothorax at 4 h on chest
radiograph. The presence of an immediate pneumothorax wasrecorded even when it represented only a very small, loculated
pocket of air. Chest radiographs taken at 4 h were carefully
reviewed for the presence of a small pneumothorax.
An immediate pneumothorax was identified in 59 of 195
patients (30%). Only 35 of 195 patients (18%) had a
pneumothorax present on chest radiograph at 4 h. One patient
had a small pneumothorax seen on chest radiograph, which was
not present on CT. This patient was asymptomatic and did not
require any treatment.Review of the case notes of all patients who sustained
pneumothoraces showed that five patients (3%) underwent
aspiration of the pneumothorax and four patients (2%) required
insertion of an intercostal drain. All patients who required
treatment were identified as having sustained pneumothoraces
on the immediate post-biopsy CT.
Only one other patient developed a significant complication in
our study; during initial infiltration of local anaesthetic, a large
haematoma developed in the subcutaneous tissues. Despitecompression this did not immediately disperse, causing local
distortion of the tissues. It was felt this would significantly hinder
needle placement and therefore the procedure was abandoned.
The patient did not suffer any lasting ill effects and underwent a
diagnostic CT-guided lung biopsy the following week.
Six patients in our study were documented to have small,
self-limiting haemoptysis after the procedure. It is possible that
other patients may have had similar minor symptoms withoutthis being documented, so no further observations about the
incidence of haemoptysis have been made. Certainly no
patients developed significant pulmonary haemorrhage after
the procedure.
The data were also analysed to assess which factors
influenced the incidence of pneumothorax post biopsy.
Initially, five factors were considered, namely the size of the
Table 1 – Final diagnoses
Final diagnosis Number of cases
MalignantSmall cell carcinoma 17Adenocarcinoma 51Squamous cell carcinoma 42Non-small cell carcinoma, not further specified 34Anaplastic carcinoma 2Malignant cell type not otherwise specified 10
BenignAbscess 1Tuberculosis 2Pneumocystis carinii pneumonia 1Infection, not further specified 3Pneumoconiosis 1Pulmonary fibrosis 1Round atelectasis 1Sarcoidosis 1Scar tissue 2Vasculitis 1Benign, not further specified 12
Table 2 – Significance of each variable on diagnostic accuracy
Characteristic Diagnostic group Non-diagnostic group p-Value
Age (years) mean ^ 2 SD 67.6 ^ 9.86 67.0 ^ 9.78 0.74Sex (m/f) 92/67 19/17 0.34
Size of lesion (mm) mean ^ 2 SD 41.4 þ /20.97 38.0 ^ 16.96 0.338Depth of lesion (mm) mean ^ 2 SD 9.8 ^ 14.28 17.2 ^ 20.7 0.054No of passes mean ^ 2 SD 1.81 ^ 0.59 1.67 ^ 0.65 0.385Needle type (FNA/core) 118/41 33/3 ,0.005Status of operator (cons/SpR) 89/70 14/22 0.15
FNA—fine needle aspirate; cons—consultant; SpR—specialist registrar.
CT-GUIDED LUNG BIOPSY: FACTORS INFLUENCING DIAGNOSTIC YIELD AND COMPLICATION RATE 793
lesion, the depth of the lesion from the pleural surface, thenumber of passes made, the type of needle used and the statusof the operator. Results are displayed in Table 3. Appropriatestatistical analysis was used to determine the significance ofeach result.
In keeping with the findings of other authors, we notedsignificant increases in the incidence of pneumothorax withincreased distance from the pleural surface and smaller lesionsize. We also showed that core biopsies were associated with alesser incidence of pneumothorax, and that the average numberof passes made was less in cases where pneumothoraxdeveloped.
DISCUSSION
Patients with intrapulmonary masses of unclear aetiologypresent clinicians with a diagnostic problem. The causes ofsuch masses encompass a wide range of both benign andmalignant processes with a diverse range of treatment options.Because of the revolution in available treatment options, it hasbecome increasingly important to reach a definitive diagnosisfor such patients in a timely fashion. Furthermore, because ofthe widespread availability of CT, increasing numbers of smallpulmonary nodules have been detected which are not visible onchest radiography, and ideally these too require definitive tissuediagnosis. Accordingly, over the past 30 years or so, themedical community has sought to modify a range of relativelynon-invasive diagnostic procedures, to improve both theiraccuracy and safety. CT-guided percutaneous transthoracicbiopsy has been at the forefront and is now accepted as being acrucial adjunct to fibre-optic bronchoscopy in the diagnosis ofpatients with suspected lung cancer.
We have observed that core biopsies have a significantlyhigher diagnostic yield than fine-needle aspirate biopsies, andthis is supported by published literature. Laurent et al. [10]compared 125 fine-needle aspirate biopsies with 98 corebiopsies; they found a significantly lower false-negative ratein the core biopsies compared with the fine-needle aspirates(2.6 and 17%, respectively). Lucidarme et al. [11] retro-spectively studied 89 patients who had undergone corebiopsies. In all cases, diagnostic material was obtained. Theyreported an overall diagnostic accuracy of 88%, subdivided into93% for malignant lesions and 71% for benign lesions. A morerecent study [12] has reported an overall diagnostic accuracy
for core biopsies of 95%. Our diagnostic rate of 93% for corebiopsies is comparable with these figures.
It has been established that core biopsies are diagnosticallymore useful than FNA biopsies in benign pulmonary disease[12,16]. Therefore, we further analysed our data excluding allbenign lesions, to confirm that our results remain significant inthe diagnosis of malignant lesions. Core biopsies remainedsignificantly more accurate than FNA biopsies when solelymalignant lesions were included ðp , 0:005Þ:
We do not report separately on the use of co-axial needlebiopsy systems. It is now established that co-axial systems havediagnostic benefits over other techniques, at least partlybecause they allow more passes to be made through the lungparenchyma [11,16]. Some of this work was published towardsthe end of our data collection period. A small number ðn ¼ 4Þof the core biopsies in our study were performed using co-axialsystems, all of which produced diagnostic results. It is now ourpolicy to routinely use co-axial systems for all core biopsies.
This study also showed that repeat biopsies, particularlyusing core needles, improve the diagnostic yield considerably.Of 12 patients undergoing repeat biopsies, 11 were accuratelydiagnosed by CT-guided biopsy. Given that other diagnosticprocedures such as video-assisted thoracoscopic biopsy andopen biopsy are more invasive, we believe repeat biopsies to beextremely useful if the first biopsy has been non-diagnostic.
The diagnostic rate we describe for FNA biopsies is lowerthan hoped. The main reason for failure to achieve diagnosiswas insufficient material, largely because it is difficult to assessFNA adequacy by visual inspection alone. A meta-analysis ofseveral previous studies [8] shows that the presence of acytopathologist at the time of FNA biopsy, with immediateassessment of the adequacy of samples, can significantlyimprove the diagnostic rate. However, like many otherinstitutions, this facility is not available to us. It is difficult tocompare this result with what is achieved elsewhere as nodirectly comparable data are published. Most studies limit theirobservations to patients with an ultimate diagnosis ofmalignancy, and quote sensitivities of 82–99% [13] in thesegroups. It has long been accepted that the yield of FNA in non-malignant lesions is much less, achieving specific benigndiagnoses in only 39–77% [14,15]. We elected to report on alldata, without selection depending on final diagnosis, as thiswould appear to mirror real practice more closely. Ultimately,in our practice, the exact benign diagnosis is less important, aswe are investigating a group of patients with high clinicalsuspicion of malignancy, and are asked to confirm or refute
Table 3 – Significance of each variable on pneumothorax rate
Characteristic Pneumothorax No pneumothorax p-Value
Age (years) mean ^ 2 SD 68.8 ^ 9.7 66.9 ^ 9.7 0.22Sex (m/f) 34/26 77/58 0.96Size of lesion (mm) mean ^ 2 SD 32.2 ^ 15.9 44.5 ^ 21.2 , 0.001Depth of lesion (mm) mean ^ 2 SD 23.7 ^ 16.3 5.7 ^ 15.8 , 0.001No of passes (mean ^ 2 SD) 1.53 ^ 0.66 1.89 ^ 0.78 0.0009Needle typeFNA 53/151 ¼ 35% 98/151 ¼ 65% 0.009Core 7/44 ¼ 16% 37/44 ¼ 84%Status of operator (cons/SpR) 31/29 72/63 0.84
FNA—fine needle aspirate; cons—consultant; SpR—specialist registrar.
CLINICAL RADIOLOGY794
this, commenting on cell type in cases where malignancy is
confirmed.
We were surprised to find that, in our study, core biopsies
were in fact associated with a lesser incidence of pneumothorax
than fine-needle aspirate samples. Further analysis of the data
shows that this is at least in part because proportionately more
pleural-based lesions were biopsied with core needles: of 102
contiguous lesions, 34 had core biopsies and 68 had FNA
biopsies. The overall incidence of pneumothorax in lesions
contiguous with the pleural surface was only 4%. The higher
proportion of core biopsies in this group is likely to explain the
apparent difference overall between incidence of pneumo-
thorax in core biopsies compared with FNA biopsies. This
finding raised the possibility that the same bias accounted for
the apparent increased diagnostic accuracy with core biopsy.
Further analysis showed that of 93 lesions non-contiguous with
the pleural surface, 83 were biopsied using FNA biopsies and
10 using core biopsies. Sixty-two (75%) of the FNA biopsies
were diagnostic, whilst 21 (25%) were non-diagnostic. In
comparison, all 10 core biopsies of lesions deep to the pleural
surface were diagnostic. This relationship tends towards
statistical significance ðp ¼ 0:07Þ despite the small numbers
in the core biopsy group. Therefore, the improved accuracy of
core biopsies compared with FNA biopsies does not appear to
be due to bias.
Interestingly, our study appears to contain a higher
percentage of lesions contiguous to the pleural surface than
might be expected: 52% of lesions biopsied were contiguous to
the pleural surface, and the average depth of all lesions from the
pleural surface was 11.1 mm. Lucidarme et al. [11] reported an
average lesional depth of 20 mm, and Lopez et al. [12] of
31 mm, while Arakawa et al. [25] reported that 31% of all
lesions biopsied were contiguous with the pleural surface. This
difference may be due to referral patterns: we are referred
patients in whom bronchoscopy has failed to provide a tissue
diagnosis, or in whom lesions have been shown be distant from
segmental bronchi, as diagnostic yield of bronchoscopy in these
patients has been shown to be reduced [26].
Previous studies [16–18] have shown that larger lesions are
biopsied with significantly higher accuracies than smaller
lesions. We did not show a significant difference in diagnostic
accuracy depending on lesion size. This would appear to be
because of an unexplained dip in diagnostic rate in lesions
measuring over 50 mm in our series: while 95% of lesions 41–
50 mm were accurately diagnosed, this fell to 80% in lesions
over 50 mm. This is at odds with previous work [16], which
showed a diagnostic rate of 100% for lesions over 50 mm in
size. Possible reasons for our results may be that larger lesions
have a higher incidence of central necrosis; this may not be
apparent on the non-contrast planning images obtained at the
start of the biopsy, leading to sampling error. Additionally,
larger lesions may be more likely to be associated with collapse
and consolidation of the surrounding lung; again this would be
difficult to delineate on non-contrast planning images, and may
be another possible source of sampling error. Other studies
have assessed the effect of size on diagnostic accuracy by
dividing the patients into two groups; Laurent et al. [17] used
20 mm as the cut-off point, whilst Li et al. [18] used 15 mm.
Analysis of our data after dividing lesions into those less
than or equal to 20 mm and those greater than 20 mm did not
show any significant difference in diagnostic yield. This may
again be because of our lower yield for the largest lesions. Our
group contained 10 lesions less than or equal to 15 mm in size;
the diagnostic rate in this group was 60% compared with 82%
in the group of lesions measuring 15 mm and above. This
difference reached statistical significance, and thus concurred
with the results found by Li et al. It is likely that biopsy of
smaller lesions is more likely to incur sampling error when
compared with larger lesions.
We did not show any improvement in diagnostic yield with
an increased number of needle passes. This is in keeping with
previous work [16,18]. Assessment of sample adequacy by
visual inspection in core biopsies could be expected to be
reasonably accurate, whereas this method is likely to be less
accurate for FNA. Therefore we analysed all FNA biopsies
separately to see if the number of passes in this group correlated
with diagnostic accuracy. No significant relationship was
found. As insufficient material for diagnostic evaluation is the
main limitation of FNA biopsies, and increasing the number of
passes does not improve diagnostic yield, it seems that, as
previously recommended, FNA biopsies should only be
performed in the presence of a cytopathologist [8], or the
failure rate because of inadequate specimens will be difficult to
improve.
We found that in patients who developed a pneumothorax
post biopsy, the average lesion size was significantly smaller
than in those who did not develop a pneumothorax. This
mirrors other recent results [13,19,20]. The reason for this
association is not entirely clear. Previous authors [21] have
postulated that smaller lesions are more difficult to biopsy, and
therefore the dwell time of the needle within the lung may be
longer; it has been suggested that movement during patient
respiration may cause tearing of the lung parenchyma and by
this mechanism longer dwell times may be the mechanism of
increased incidence of pneumothorax. However, Ko et al. [22]
recently published data showing that an increased needle dwell
time was not correlated with incidence of pneumothorax, thus
apparently refuting this suggested mechanism. It could be
postulated that larger lesions are more likely to be adjacent to
the pleural surface, and for this reason are associated with a
lower risk of pneumothorax. However, our data for lesion size
are highly significant and remain so when lesions contiguous to
the pleural surface are excluded from analysis. Other authors
[13] have also shown this. In our opinion, the relationship
between lesion size and incidence of pneumothorax may be
explained by the procedural technique. The throw of automated
core biopsy needles in widespread use is usually 2 cm; this may
mean that in the biopsy of smaller lesions, the throw of the
needle extends beyond the boundary of the lesion and into the
lung parenchyma. It seems likely that traversing normal lung in
this way may increase the pneumothorax rate. Additionally, the
technique used in FNA biopsies of moving the needle tip back
and forwards within the lesion several times may also involve
the needle track extending into the lung parenchyma during the
biopsy of smaller lesions and thus increase the incidence of
pneumothorax. Certainly, the relationship of increasing pneu-
mothorax incidence with decreasing lesion size is reasonably
well established.
CT-GUIDED LUNG BIOPSY: FACTORS INFLUENCING DIAGNOSTIC YIELD AND COMPLICATION RATE 795
The relationship between increasing lesion depth from the
pleural surface and increasing incidence of pneumothorax
was highly significant in our study. Patients developing
pneumothorax had an average lesion depth of 24 mm compared
with 6 mm in those who did not develop pneumothorax. This is
in keeping with previous results [13,19,20]. The most obvious
reason for these findings is the much lower incidence of
pneumothorax after biopsy of lesions contiguous to the pleural
surface; in our study only 4% of the 102 contiguous lesions
developed pneumothorax. The incidence increased to 62% for
all lesions lying deep to the pleura. When the data were further
analysed to see if depth was correlated with incidence of
pneumothorax when all contiguous lesions were excluded, no
significant relationship was found. Therefore we conclude that,
as might be expected, CT-guided biopsies in which aerated
lung is not traversed are associated with a very low incidence of
pneumothorax; as soon as aerated lung is traversed, the risk
rises considerably, but thereafter the exact depth of the lesion is
not a significant factor.
Previous work has not shown any relationship between
the number of pleural passes made and the incidence of
pneumothorax [20,21,23]. Unexpectedly, we showed that the
average number of pleural passes was in fact less in patients
developing a pneumothorax. Presumably, the explanation for
this is that when a pneumothorax occurred on the first
pleural pass, the procedure was terminated.
Our documented incidence of intervention falls within the
lower end of reported rates of intercostal drain insertion,
which range from 0% [18] to 14.9% [20]. Previous studies
do not describe pleural aspiration as a method of treating
post-procedure pneumothorax; we are not responsible for
treatment decisions in our patients, however it is of note that
all five patients who had pleural aspiration were successfully
treated, with no requirement for subsequent intercostal drain
insertion.
The data on the nine patients in our series who required
treatment for pneumothorax were separately reviewed and we
found no common features that could be used to predict this
outcome. It has been shown before that of the patients who
develop pneumothorax, those with chronic obstructive airways
disease have a higher incidence of necessity of intercostal drain
insertion [24]. These data were not available for a proportion of
our patients, therefore we do not report on these. The role of
abnormal lung function as a risk factor for pneumothorax
development post biopsy has been fiercely debated in the
literature [19–23]. No consistent conclusions have been
reached, and the consensus reached is that abnormal lung
function probably does not increase the risk of pneumothorax,
although if a pneumothorax occurs, the incidence of require-
ment for chest drain insertion is higher. No other factors that
can predict requirement for chest drain insertion have been
reported in the literature.
Our findings suggest that core biopsy should be the method
of choice in CT-guided lung biopsy, especially if a cytopathol-
ogist is not available at the time of the procedure. We expect
that accuracies of 90% and over should be achieved, with no
excess of complications. Diagnostic yield should be further
improved by repeat CT biopsy when the first biopsy is non-
diagnostic. The incidence of pneumothorax appears dependent
on the position of the lesion in question; operator variablesappear less important. Should pneumothorax occur, pleuralaspiration may be an appropriate therapeutic measure.
Acknowledgements. We are grateful to Dr Kendra Murray, ConsultantPathologist, for her help in the acquisition of pathological results, and to DrJoanne Moncrieff for statistical advice.
REFERENCES
1 Haaga JR, Alfidi RJ. Precise biopsy localisation by computedtomography. Radiology, 1976;118:603–607.
2 Tsukada H, Satou T, Iwashima A, Souma T. Diagnostic accuracy of CTguided automated needle biopsy of lung nodules. AJR Am J Roentgenol,2000;175:239–243.
3 Li H, Boiselle PM, Shepard JO, et al. Diagnostic accuracy and safety ofCT-guided percutaneous needle aspiration biopsy of the lung:comparison of small and large pulmonary nodules. AJR Am JRoentgenol, 1996;167:105–109.
4 van Sonneberg E, Casola G, Ho M. Difficult thoracic lesions: CT guidedbiopsy experience in 150 cases. Radiology, 1988;167:457–461.
5 Kazerooni EA, Lim FT, Mikhail A, Martinez FJ. Risk of pneumothoraxin CT-guided transthoracic needle aspiration biopsy of the lung.Radiology, 1996;198:371–375.
6 Norenberg R, Claxton CP, Takaro T. Percutaneous needle biopsy of thelung. Report of two fatal complications. Chest, 1974;66:216–218.
7 McCartney R. Haemorrhage following percutaneous lung biopsy.Radiology, 1974;112:305–307.
8 Austin JH, Cohen MB. Value of having a cytopathologist present duringfine needle aspiration biopsy of lung: report of 55 cancer patients andmeta analysis of the literature. AJR Am J Roentgenol, 1993;160:175–177.
9 McLoud TC. Should cutting needles replace needle aspiration of lunglesions. Radiology, 1998;208:569–570.
10 Laurent F, Latrabe V, Vergier B, Michel P. Percutaneous CT-guidedbiopsy of lung: comparison between aspiration and automated cuttingneedles using a coaxial technique. Cardiovasc Intervent Radiol, 2000;23:266–272.
11 Lucidarme O, Howarth N, Finet J, Grenier P. Intrapulmonary lesions:percutaneous automated biopsy with a detachable, 18 gauge, coaxialcutting needle. Radiology, 1998;207:759–765.
12 Lopez H, Vogl T, Ricke J, Felix R. CT-guided percutaneous corebiopsies of pulmonary lesions. Diagnostic accuracy, complications andtherapeutic impact. Acta Radiol, 2001;42:151–152.
13 Cox J, Chiles C, McManus C, et al. Transthoracic needle aspirationbiopsy: variables that affect risk of pneumothorax. Radiology, 1999;212:165–168.
14 Stanely JH, Fish GD, Andriole JG. Lung lesion: cytologic diagnosis byfine needle biopsy. Radiology, 1987;172:389–391.
15 Klein JS, Salomon G, Stewart F. Transthoracic needle biopsy with acoaxially placed 20-gauge automated cutting needle: results in 122patients. Radiology, 1996;198:715–720.
16 Tsukada H, Satou T, Iwashima A, Souma T. Diagnostic accuracy of CTguided automated needle biopsy of lung nodules. AJR Am J Roentgenol,2000;175:239–243.
17 Laurent F, Latrabe V, Vergier B, et al. CT-guided transthoracic needlebiopsy of pulmonary nodules smaller than 20 mm: results with anautomated 20-gauge coaxial cutting needle. Clin Radiol, 2000;55:281–287.
18 Li H, Boiselle PM, Shepard JO, et al. Diagnostic accuracy and safety ofCT-guided percutaneous needle aspiration biopsy of the lung:comparison of small and large pulmonary nodules. AJR Am JRoentgenol, 1996;167:105–109.
19 Laurent F, Michel P, Latrabe V, et al. Pneumothoraces and chest tubeplacement after CT-guided transthoracic lung biopsy using a coaxialtechnique. AJR Am J Roentgenol, 1999;172:1049–1053.
20 Kazerooni EA, Lim FT, Mikhail A, Martinez FJ. Risk of pneumothoraxin CT-guided transthoracic needle aspiration biopsy of the lung.Radiology, 1996;198:371–375.
CLINICAL RADIOLOGY796
21 Miller KS, Fish GB, Stanely JH, Schabel SI. Prediction of pneumo-thorax rate in percutaneous needle aspiration of the lung. Chest, 1988;93:742–745.
22 Ko J, Shepard J, Drucker E, et al. Factors influencing pneumothoraxrate at lung biopsy: are dwell times and angle of pleural puncturecontributing factors? Radiology, 2001;218:491–496.
23 Poe RH, Kallay MC, Wicks CM, Odoroff CL. Predicting the risk ofpneumothorax in needle biopsy of the lung. Chest, 1984;85:232–235.
24 Anderson CL, Crespo JC, Lie TH. Risk of pneumothorax not increasedby obstructive lung disease in percutaneous needle biopsy. Chest, 1994;105:1705–1708.
25 Arakawa H, Nakajima Y, Kurihara Y, et al. CT-guided transthoracicneedle biopsy: a comparison between automated biopsy gun and fineneedle aspiration. Clin Radiol, 1996;51:503–506.
26 Wong PC, Lee J, Lam FM, et al. Fibreoptic bronchoscopy in thediagnosis of lung cancer. Monaldi Arch Chest Dis, 1999;54:394–398.
CT-GUIDED LUNG BIOPSY: FACTORS INFLUENCING DIAGNOSTIC YIELD AND COMPLICATION RATE 797