Musculoskeletal Imaging
Original Research
Percutaneous CT-Guided Bone Biopsy: Diagnosis of Malignancy in Lesions With Initially Indeterminate Biopsy Results and CT Features Associated With Diagnostic or Indeterminate Results
Audio Available | 
Share
OBJECTIVE. The purpose of our study was to determine the proportion of bone lesions with indeterminate results after initial percutaneous CT-guided bone biopsy that ultimately are found to be malignant and whether CT features are associated with diagnostic outcomes.
MATERIALS AND METHODS. The results of 800 consecutive percutaneous CT-guided bone biopsies performed at a tertiary cancer center were reviewed. The initial histopathologic diagnosis was classified as diagnostic or indeterminate. On the basis of follow-up information, indeterminate results were subcategorized as benign, malignant, or persistently indeterminate. Two readers independently analyzed the CT images.
RESULTS. Initial percutaneous CT-guided bone biopsy was diagnostic in 69% and indeterminate in 31%. Malignancy was diagnosed in 90% of initially diagnostic results. In lesions with initially indeterminate results, a diagnosis was subsequently made in 62%; 39% of subsequent diagnoses were malignant as of the last available follow-up. CT features associated with diagnostic results included cortical destruction and large extraosseous mass (p < 0.05). More lesional sclerosis and presence of fat were associated with indeterminate results (p < 0.001). CT features associated with malignant results included less-extensive sclerosis and lesser sclerotic rim (p < 0.05). Increased age, female sex, and a cancer history were associated with higher risk of malignancy among patients with diagnostic results at initial biopsy.
CONCLUSION. Bone lesions that initially yield indeterminate results at percutaneous CT-guided bone biopsy often are subsequently shown to be malignant; vigorous pursuit of a diagnosis is recommended if initial results are indeterminate. Lesions showing fat or more sclerosis are more likely to be indeterminate; lesions with less sclerosis or smaller sclerotic rim are more likely to yield malignant results.
Keywords: bone lesion, CT-guided bone biopsy, indeterminate bone lesion, malignancy, osseous metastases
Assessment of bone lesions is a common and essential activity in the care of oncology patients, and percutaneous CT-guided bone biopsy is a minimally invasive and safe biopsy technique in the assessment of bone lesions. The diagnostic yield often exceeds 70% and has been comparable among various studies in the English language radiology literature [1–4]. However, when the results of percutaneous CT-guided bone biopsy are indeterminate, cancer staging and therapy planning can be compromised, delaying treatment and possibly decreasing the chances for survival and cure. Indeterminate results, such as “no tumor cells,” can be related to many factors, including inadequate biopsy technique, or to a specific pathologic process that is difficult to diagnose on the basis of a small amount of tissue. Yet it is not prudent to consider such a result benign or as caused by metastasis from a known cancer; a recent study of 93 patients with a single known malignancy found that percutaneous CT-guided bone biopsy showed a second primary neoplasm in 8% of patients [5]. Recognition of various CT features associated with diagnostic and indeterminate percutaneous CT-guided bone biopsy results may be useful in predicting the expected diagnostic yield and thus in helping to plan the most effective biopsy method (imaging-guided vs open surgical biopsy) and to select a specific bone lesion that is most likely to yield a diagnostic result when multiple bone lesions are present.
The purpose of this study is to determine the diagnostic yield of percutaneous CT-guided bone biopsy, the proportion of bone lesions with indeterminate results after initial percutaneous CT-guided bone biopsy that ultimately were found to be malignant, and whether any CT features are associated with diagnostic outcomes.
| Materials and Methods |   |
|---|
This retrospective study was approved by our institutional review board, which waived the need for informed consent. A computerized search of the institutional tumor registry and radiology database identified 800 patients who underwent percutaneous CT-guided bone biopsy between January 2000 and December 2006 at our tertiary cancer center. The study population consisted of 456 females (57%) and 344 males (43%), with a median age of 60 years (age range, 3–91 years). The pelvis (n = 225), lumbar spine (n = 140), sacrum (n = 125), thoracic spine (n = 127), and ribs (n = 63) accounted for 85% of all biopsy sites; other less common sites included femur (n = 32), sternum (n = 25), scapula (n = 21), clavicle (n = 12), cervical spine (n = 9), humerus (n = 6), and other distal bones and skull (n = 15).
The institutional electronic medical record for each patient was reviewed for any history of cancer before percutaneous CT-guided bone biopsy, histopathologic diagnosis at percutaneous CT-guided bone biopsy, clinical and imaging follow-up, and any subsequent surgical biopsy results. On the basis of clinical outcome determined from subsequent imaging follow-up, clinic notes, or surgical biopsy results, initially indeterminate percutaneous CT-guided bone biopsy results were subclassified as benign, indeterminate, or malignant. A benign result was defined as a surgical biopsy that revealed a benign process or the lesion was clearly benign at follow-up imaging (such as healing of a fracture or no change in size at imaging for more than 3 years). An indeterminate result was defined as imaging and clinical follow-up for less than 3 years without any specific diagnosis being achieved. A malignant result was defined as a surgical biopsy that revealed malignancy or subsequent imaging studies that showed features compatible with malignancy. Such features are those used routinely in daily clinical oncologic practice. The most common feature encountered was enlargement or increased lysis of the biopsied lesion on serial imaging studies in a patient with a known primary malignancy. Another compelling feature was the presence of additional similar-appearing bone lesions. Radiologic evidence of healing (i.e., reduction in size or development of intralesional sclerosis without lesion enlargement) after chemotherapy or radiation therapy was considered consistent with malignancy. Several of the lesions were deemed malignant because they showed increased activity at baseline PET that resolved after therapy. None of those patients had clinical evidence to suggest that an enlarging bone lesion was caused by infection.
Baseline CT examination was defined as the diagnostic CT (n = 579) or PET/CT examinations (n = 59) obtained within 2 months before percutaneous CT-guided bone biopsy or the CT images obtained during bone biopsy when no such diagnostic CT or PET/CT examination was available (n = 162). Two musculoskeletal tumor radiologists with 4 and 10 years of experience independently reviewed the baseline CT examinations on a PACS workstation (Centricity, GE Healthcare) using soft-tissue and bone window settings. The readers were blinded to the cancer history of the patients and any histologic diagnoses of bone lesions. The following CT features of each lesion were recorded: matrix (relative fractions of lysis and sclerosis), attenuation, border (definable, sclerotic), size, presence of cortical involvement (destruction, thickening, periosteal reaction), presence of fluid-fluid levels, and associated extraosseous mass. Matrix in a bone lesion was classified as lytic or sclerotic via qualitative comparison with adjacent normal bone. The readers scored the relative amounts of matrix, definable border, and sclerotic rim as a fraction using the scale of none of the lesion, less than one third of the lesion, one third to two thirds of the lesion, more than two thirds of the lesion, or throughout the entire lesion. Attenuation was categorized as fat (< 0 HU), fluid (0–20 HU), or soft tissue (> 20 HU). The longest dimensions of the intraosseous component and of any associated extraosseous mass in the axial plane were recorded. The location of the biopsy needle (in bone vs associated extraosseous mass) was recorded.
Percutaneous CT-guided bone biopsy was performed by several interventional radiologists. All patients received moderate sedation using IV midazolam and fentanyl with continuous monitoring of vital signs by a sedation nurse. The skin overlying the biopsy needle trajectory was aseptically prepared and locally anesthetized. Needle size ranged from 22 to 11 gauge, and needles from various manufacturers were used, including Westcott (BD Medical), Bonopty (AprioMed), Ostycut (Angiomed), Percucut (E-Z-EM), and Laurane (Laurane Medical).
A standard coaxial technique was used. The number of tissue sampling passes was not documented in reports. Fine-needle aspiration (FNA) or core specimens were obtained at the discretion of the operators. For certain indications, such as infection or documentation of metastatic disease when the overlying cortex was disrupted, FNA samples were commonly obtained. Otherwise, “visually adequate” core specimens were obtained. A cytopathologist evaluated all specimens onsite, and core specimens were sent to surgical pathology. More specimens were obtained until diagnostic adequacy was confirmed by the cytopathologist onsite or the operator determined that the biopsy lesion was adequately sampled. Additional CT images were obtained at the conclusion of biopsy to assess for immediate complications.
Assessment of interreader agreement— Interreader agreement on imaging features was assessed using the kappa statistic for binary variables, the Fleiss-Cohen weighted kappa statistic for categoric variables, and the concordance cor relation coefficient [6] for continuous variables.
Association between imaging features and diagnostic outcome—On the basis of initial percutaneous CT-guided bone biopsy results, we performed analyses of CT imaging features to differentiate indeterminate outcomes from diagnostic outcomes and malignant lesions from benign lesions using univariate and multivariable logistic regression. Because the likelihoods of indeterminate versus diagnostic and malignant versus benign lesions could be affected by age, sex, and cancer history, the multivariate analyses were adjusted for differences in these characteristics. Retaining these characteristics, the multivariate model used stepwise selection to identify statistically significant CT imaging features with the entry significance level at 0.10 and the stay significant level at 0.05. Some lesions were not seen by reader A (71 lesions) or reader B (25 lesions). Many of the lesions not visualized by the readers were rather subtle at CT. Even though the readers were not blinded to the location of the lesion (because they saw on the images where the CT biopsy needle had been placed), they attempted to assess the bone in that region for abnormal findings. Needle placement at percutaneous CT-guided bone biopsy in such cases often depended on correlation with findings visible at MRI. Therefore, the analyses were performed for each reader separately.
The odds ratio (OR) along with the 95% CI were estimated in the logistic regression. An OR greater than 1 indicated a positive association between the CT feature and indeterminate biopsy when differentiating indeterminate from diagnostic results or a positive associate between the CT feature and malignant biopsy outcomes when differentiating initial malignant from benign lesions. An OR less than 1 indicated a negative association accordingly. A test with the p value less than 0.05 was considered statistically significant. All analyses were conducted in SAS version 9.2 statistical software (SAS Institute).
| Results | |
|---|
Six hundred thirty-nine (80%) of 800 patients had a known diagnosis of cancer before percutaneous CT-guided bone biopsy. Four hundred forty-one (80%) of 551 patients with initially diagnostic results and 198 (80%) of 249 patients with initially indeterminate results had at least one known cancer. Among those with initially diagnostic results, 411 (83%) of 495 patients with malignant results and 30 (54%) of 56 patients with benign results had one or more known cancers.
 View larger version (16K) | Fig. 1 —Flowchart shows results of percutaneous CT-guided bone biopsy in 800 patients. |
TABLE 1: Performance of Percutaneous CT-Guided Bone Biopsy in 800 Patients
The diagnoses included cancer of the breast (n = 271), prostate (n = 76), lung (n = 51), lymphoma (n = 44), colorectum (n = 40), urinary tract (n = 32), head and neck (n = 28), melanoma (n = 15), multiple myeloma (n = 15), thyroid (n = 14), kidney (n = 13), leukemia (n = 13), soft-tissue sarcoma (n = 12), ovary (n = 10), uterus (n = 10), Ewing sarcoma (n = 8), Kaposi sarcoma (n = 7), pancreas (n = 7), esophagus (n = 5), skin (n = 5), unknown primary carcinoma (n = 5), small bowel (n = 3), osteosarcoma (n = 3), stomach (n = 3), chondrosarcoma (n = 2), chordoma (n = 2), hepatocellular carcinoma (n = 2), plasma cell neoplasm (n = 2), brain (n = 1), cervix (n = 1), myelodysplastic syndrome (n = 1), mesothelioma (n = 1), testes (n = 1), thymoma (n = 1), and vagina (n = 1). The numbers of diagnoses add to more than the number of patients because some patients had more than one known cancer.
The percutaneous CT-guided bone biopsy was diagnostic in 69% of lesions and indeterminate in 31% (Fig. 1). Percutaneous CT-guided bone biopsy was performed in nearly equal numbers of axial and appendicular skeletal sites (Table 1). The most common biopsy site in the axial skeleton was the lumbar spine (35%) and in the appendicular skeleton was the pelvis (56%). The diagnostic yield was higher in axial lesions (73%) than in appendicular lesions (65%). The diagnostic yield in axial lesions was highest in the cervical spine (78%), and lowest in the lumbar spine (69%). The ribs were the appendicular site with the highest yield (70%); the humerus had the lowest yield (50%).
Among diagnostic results, malignancy accounted for 495 (90%) of 551 lesions: 260 (89%) of 292 axial lesions and 235 (91%) of 259 appendicular lesions. Breast cancer bone metastases were the most common malignancy, diagnosed in 168 (34%) of 495 malignant lesions. Multiple myeloma was the most common primary bone malignancy, diagnosed in 27 lesions (5%) (Fig. 2). The most common benign diagnosis at percutaneous CT-guided bone biopsy was osteomyelitis, followed by hemangioma and osteonecrosis (Fig. 3). Diagnosis was subsequently made in 154 (62%) of 249 initially indeterminate bone lesions at percutaneous CT-guided bone biopsy on the basis of subsequent imaging studies and clinical notes (n = 114 lesions) and surgical biopsy (n = 40 lesions).
In patients with initially indeterminate bone lesions, 60 (39%) of 154 bone lesions subsequently were shown to be malignant and 94 (61%) were categorized as benign (Fig. 1). Diagnosis persistently remained indeterminate in 95 (38%) of 249 initially indeterminate lesions because of incomplete follow-up and lack of surgical biopsy. The median follow-up of initially indeterminate results was 39 months (range, 0.3–133 months).
 View larger version (12K) | Fig. 2 —Graph shows ten most common types and frequencies of malignancy diagnosed at initial percutaneous CT-guided bone biopsy. |
 View larger version (14K) | Fig. 3 —Graph shows types and frequencies of benign diagnoses at initial percutaneous CT-guided bone biopsy. Other benign tumors include enchondroma, osteochondroma, osteoblastoma, and desmoplastic fibroma. |
Among the 40 lesions that underwent subsequent surgical biopsy, 11 were malignant and 29 were benign. The most common malignant diagnosis was bone metastases (5/11) from cancers of the breast, skin, prostate, lung, and colon; the other malignant diagnoses included various primary bone malignancies (chondrosarcoma [n = 2] and Hodgkin disease, radiation-related sarcoma, odontogenic tumor, and plasma cell with amyloid deposit [n = 1 each]). The most common benign surgical diagnosis was hemangioma (4/29); other benign diagnoses included fibrous dysplasia, chondromyxoid fibroma, inflammation, acellular bone marrow, or granulation tissue (n = 3 each); osteonecrosis or fibrosis (n = 2 each); bone healing, osteomyelitis, desmoplastic fibroma, Paget disease, schwannoma, or foreign body reaction (n = 1 each).
Eight of 800 lesions underwent repeat percutaneous CT-guided bone biopsy. Initial biopsy results consisted of one schwannoma, one suspected lymphoma, and six indeterminate lesions. At second percutaneous CT-guided bone biopsy, one benign lesion was confirmed to be schwannoma and one lesion with a suspected diagnosis of lymphoma was indeterminate and later diagnosed as myelodysplastic syndrome at clinical follow-up.
Six initially indeterminate lesions remained indeterminate at second percutaneous CT-guided bone biopsy. Diagnoses were made subsequently in three of these six indeterminate lesions, and the other three lesions were lost to follow-up. Two lesions were determined to be benign (enchondroma by surgical biopsy and benign lesion by imaging), and one lesion was determined to be a breast cancer metastasis on the basis of its serial growth at CT.
Soft-tissue attenuation and cortical destruction were the most common CT features noted by both readers (Table 2). Fat attenuation was a less common feature, recorded in 4% of lesions by reader A and 7% by reader B. The most common CT matrix was mixed sclerotic and lytic (one third to two thirds) by reader A and more lytic (more than two thirds) by reader B. The median size of intraosseous bone lesions was 2.6 cm for reader A and 3.5 cm for reader B (Table 2). The median size of extraosseous masses was 2.4 cm for reader A and 2.8 cm for reader B (Table 2).
Agreement between the two readers was substantial (κ ≥ 0.7) for lesion matrix, soft-tissue attenuation, and size of extraosseous mass; agreement was moderate (κ ≥ 0.4) for sclerotic rim, size of intraosseous lesion, fat attenuation, and cortical destruction (Table 2). The interreader agreement was lowest for well-defined border, fluid attenuation, cortical thickening, and periosteal reaction (Table 2). Only one bone lesion had fluid-fluid levels seen by both readers; the final diagnosis was aneurysmal bone cyst.
The analysis of the association between CT features and indeterminate biopsy at initial percutaneous CT-guided bone biopsy is summarized in Table 3. Except for fluid attenuation, fluid-fluid levels, and periosteal reaction, all the other CT features from at least one reader were associated with diagnostic or indeterminate results (p < 0.05) at univariate analysis.
At multivariate analysis (Table 3), fat attenuation and sclerotic matrix showed statistically significant associations with indeterminate biopsy results (p < 0.001 for both readers). The likelihood of bone lesions with fat attenuation being indeterminate at percutaneous CT-guided bone biopsy was 4.9–7.8 times higher than those without fat attenuation. For each increment in extent of sclerosis (0, < one third, one third to two thirds, > two thirds, or complete), the likelihood of a bone lesion being indeterminate at percutaneous CT-guided bone biopsy increased 1.4–1.5 times. For completely sclerotic lesions, percutaneous CT-guided bone biopsy results were indeterminate in 55–56% of cases. Presence of cortical thickening as noted by reader A was 2.3 times as likely to be associated with indeterminate results, and each increment in extent of a well-defined border noted by reader B increased the likelihood of indeterminate results 1.2 times.
TABLE 2: Summary of CT Features
The presence of cortical destruction and a larger size of an associated extraosseous mass each showed a negative association with indeterminate percutaneous CT-guided bone biopsy results (Table 3). Cortical destruction resulted in an OR of 0.4–0.5 for a lesion being indeterminate. For every 1-cm increment in size of an associated extraosseous mass, the OR of the lesion being indeterminate was 0.7–0.8.
The analysis of the association between CT features and malignant results at initial percutaneous CT-guided bone biopsy is summarized in Table 4. At univariate analysis, fat attenuation, cortical thickening, larger fraction of sclerosis, well-circumscribed border, and sclerotic rim were less likely to be associated with malignant results independently by each reader (OR < 1, p < 0.05). When adjusted for age, sex, and cancer history at multivariate analysis, a larger fraction of sclerosis and sclerotic rim showed negative associations with malignant results (OR < 1, p < 0.05). Neither the size of an intraosseous lesion nor the size of an associated extraosseous mass was associated with malignant results at percutaneous CT-guided bone biopsy.
Cancer history was not univariately associated with biopsy outcomes. However, at multivariate analysis (Table 3), the OR for having indeterminate results in patients with a cancer history was 0.5, meaning that a previous cancer history increases the likelihood of diagnostic results by a factor of two. Age, sex, or needle location within bone lesions (intraosseous vs soft tissue) did not show statistically significant associations with diagnostic results (p > 0.05). However, increased age, female sex, and cancer history showed statistically significant associations with malignant biopsy results (p < 0.05) (Table 4).
| Discussion | |
|---|
The diagnostic yield at percutaneous CT-guided bone biopsy is influenced by many factors, including the specific underlying pathology, lesion morphology at imaging, particular skeletal site involved, patient demographics, imaging modality and technique used during biopsy, and operator experience [1–4]. Multiple previous series on imaging-guided bone biopsy have reported variable diagnostic yields (61–91%) and described associations between multiple factors and subsequent biopsy results [1–3, 7–10]. Our study showed a comparable diagnostic yield (69%) and confirmed statistically significant associations between various CT features and diagnostic yields. The rate of indeterminate results was 31%, which is near the high end of the 8–33% range reported in other studies [11, 12].
The association between CT features of a lesion and diagnostic yield at percutaneous CT-guided bone biopsy has been reported previously, such as a decreased likelihood of obtaining a diagnosis in sclerotic lesions [1–4]. However, the CT image analysis in most previous series was performed in consensus by the readers, limiting the ability to generalize the association between CT features and diagnostic yield. To our knowledge, this is the first such study in which various CT features were independently analyzed by two bone tumor radiologists in a large patient population.
TABLE 3: Association of CT Features and Indeterminate Results at Initial Percutaneous CT-Guided Bone Biopsy
Most initially diagnostic percutaneous CT-guided bone biopsy results were bone metastases, which can be related to a known primary cancer or to a second primary cancer. Percutaneous CT-guided bone biopsy helps to distinguish the two. In subsequent diagnoses of bone lesions that were initially indeterminate at percutaneous CT-guided bone biopsy, the proportion of malignancy was 39% as of last available follow-up, substantially higher than the 10% and 24% in other studies [1, 3]. This higher proportion of malignancy and bone metastases at initial and subsequent diagnoses is probably due to our inclusion of a high proportion of patients with known cancer, who have a higher risk of developing bone metastases; our reliance on percutaneous CT-guided bone biopsy to diagnose bone metastases for staging; and our clinicians’ preference of surgical biopsy of primary bone tumors rather than percutaneous CT-guided bone biopsy. At surgical biopsy, benign diagnoses (73%) were far more common than malignant diagnoses (27%), likely because of the difficulty in making a definitive benign diagnosis on the basis of a small amount of tissue that may be insufficient for bone matrix and lesion architecture evaluation and for cytogenetic analysis [1, 13]. Previous reports of lower diagnostic accuracy in benign bone tumors than in malignant bone tumors also probably reflect this diagnostic challenge [11, 12].
TABLE 4: Association Between CT Features and Malignant Results at Initial Percutaneous CT-Guided Bone Biopsy
Axial lesions (73%) were more likely to yield diagnostic results than were appendicular lesions (65%). This is possibly due to the tendency of bone metastases to involve areas of red marrow, which are more abundant in the axial skeleton [14]. The diagnostic yield in rib lesions (70%) was the highest among the appendicular sites, even though rib lesions are considered difficult to biopsy because of the narrow bone diameter and breathing-related rib motion during biopsy. The diagnostic yield at repeat percutaneous CT-guided bone biopsy was 13%, substantially lower than the 83% obtained in a recent larger study [3]. Our study may have contained too few such cases of repeat percutaneous CT-guided bone biopsy from which to draw a meaningful comparison.
At univariate analysis, age, sex, and cancer history were not associated with diagnostic outcomes at percutaneous CT-guided bone biopsy. However, in multivariate analysis, the association became significant after adjusting for CT features, including cortical destruction, sclerosis, size, and extraosseous mass. Patients with cancer history were more likely to have diagnostic results at percutaneous CT-guided bone biopsy given the same CT features. Increased age, female sex, and cancer history at multivariate analysis adjusted for CT features were associated with increased risk of malignant diagnoses (Tables 3 and 4). In particular, female sex and cancer history independently increased the risk of malignant diagnoses by three- to fourfold, given the same age and CT features.
In our multivariate analysis, CT features that showed statistically significant associations with diagnostic results were the presence of cortical destruction and a larger size of an associated extraosseous mass. CT features that showed statistically significant associations with indeterminate results were fat attenuation and sclerosis. These CT features showed moderate and substantial interreader agreement (κ = 0.44–0.76) and may be reliably used in determining the likelihood of diagnostic outcomes in percutaneous CT-guided bone biopsy, independent of the individual readers.
Although fat attenuation in lesions was uncommon in our series, bone lesions with fat attenuation were 5–8 times as likely to have indeterminate results. This substantial association between fat attenuation and indeterminate results may be related to the fact that most fat-containing bone lesions are benign, with the associated increased difficulty in making a benign diagnosis from small tissue samples [1, 13, 15]. In a study of 184 bone tumors at MRI by Simpfendorfer et al. [15], benign tumors were much more likely to contain fat (23%) than were malignant tumors (4%).
A larger fraction of intralesional sclerosis was associated with a higher rate of indeterminate results at percutaneous CT-guided bone biopsy. Although some studies reported a diagnostic yield comparable to that in lytic lesions [16, 17], the negative association between lesional sclerosis and diagnostic yield, observed by both readers in our study, was also previously reported in other studies [1, 3, 4, 18, 19]. For each one third increment in the relative fraction of sclerosis in our study, the likelihood of indeterminate results increased 1.4–1.5 times. This decrease in diagnostic yield in sclerotic lesions is possibly due to the relatively lower tumor cellularity in sclerotic lesions and the presence of reactive bone formation, which make diagnosis even more challenging within biopsy specimens that are more susceptible to damage during processing [1, 18, 20, 21].
On the other hand, cortical destruction and extraosseous mass were more likely associated with diagnostic results, as also noted in previous studies [1–3]. These features are hallmarks of aggressive processes, whether benign or malignant in nature [22–24]. Aggressive tumors tend to show increased cellularity and a higher histologic grade, which are readily recognizable features that likely facilitate their histologic diagnosis even when the biopsy specimen is small.
Interestingly, our study found that the size of an extraosseous mass but not the size of the intraosseous lesion was a statistically significant predictor of a diagnostic outcome at multivariate analysis. We postulate that a larger extraosseous mass reflects a more aggressive process, and tissue specimens from outside a bone facilitate diagnosis without interference from bone artifacts seen in intraosseous lesions. Furthermore, because lesions were measured only in the axial plane, the possible variations in size of intraosseous components were limited.
A negative association between lesional sclerosis and a malignant diagnosis was observed independently for both readers. For each increment in the fraction of sclerosis, bone lesions were 0.6–0.7 times as likely to be diagnosed as malignant. This is partially because of the increased diagnostic difficulty in sclerotic lesions, resulting in a higher rate of indeterminate results [1, 3, 4, 18, 19]. Another possible explanation is that, except for a minority of primary malignancies such as osteosarcoma, osteoblastic metastases are less common than osteolytic metastases [25], and fewer biopsied lesions containing sclerosis are metastases. An additional compounding factor in our study was that some cancer patients were probably on systemic therapy for visceral metastases, and bone metastases became apparent only after sclerosis occurred as a healing process [26]. Fewer malignant cells may have been present, making a diagnosis even more challenging in a small specimen.
Another CT feature that showed a negative association with malignancy was a sclerotic rim. For each increment in relative fraction of sclerotic rim, a bone lesion was 0.4–0.5 times as likely to be diagnosed as a malignant lesion. In bone tumors and tumor-like lesions, the nature of the margin of a lesion is well known to correlate with lesion aggressiveness. The presence of a sclerotic rim is suggestive of a nonaggressive and benign process, such as nonossifying fibroma or intraosseous lipoma [27]. A sclerotic rim in a bone metastasis at initial staging is considered rare (1–3%) [14, 28]. However, the emergence of a sclerotic rim has been observed in healing of lytic bone metastases, such as those from breast and thyroid cancers, and is considered a sign of a partial therapeutic response [29–31]. Therefore, when a sclerotic rim associated with a bone lesion is present before percutaneous CT-guided bone biopsy, a careful correlation with history of systemic therapy and comparison with prior CT are important to avoid biopsying benign lesions or healing metastases.
Our study had several limitations, some inherent to the retrospective study design. The biopsy technique was variable depending on differences in experience of the biopsy operators, size of biopsy needles selected, and number of needle passes used to obtain tissue specimens. The pathologists also had different levels of experience, and some were not subspecialized in bone pathology. The CT technique was not uniform with regard to parameters, such as slice thickness and field of view. Clinical follow-up was incomplete in some patients because some were lost to follow-up or relevant information was not available in the electronic medical record. The referral pattern to our tertiary cancer center introduces biases; however, our results could be generalizable to populations at other tertiary cancer centers. The fact that primary bone tumors at our institution undergo open surgical biopsy rather than percutaneous CT-guided bone biopsy introduces a case selection bias. Nevertheless, our results are generalizable to patients with nonbone primary tumors. Information from radiographs, MR images, and bone scans also is used to assess a bone lesion in the clinical setting before biopsy; thus, the decision to analyze only the CT images is somewhat artificial. However, CT represents a common means for localizing such lesions for biopsy and provides often underutilized information about the lesion.
Some lesions were not seen by reader A (8.9%) or reader B (3.1%). Although the exact reasons for this discrepancy are unclear, we postulate several reasons. The location of a biopsy needle in a bone was evident on the images reviewed, leading to unavoidable bias in lesion selection by the reader, which in turn may have resulted in overinterpretation of normal bone as a lesion. Differences in reader experience and threshold for diagnosing a bone lesion, such as mischaracterizing a subtle bone lesion as focal osteopenia, also may have played a role. Some lesions were subtle or located away from the needle, which thereby distracted readers from the actual lesion. Many of the CT features of the bone lesions showed high interobserver variability, which might result in underestimation of the association with malignant or indeterminate results. Additional studies with other readers might yield different results. The interreader agreement for features, such as fat or fluid attenuation, was less than might be expected. In such cases, the amount of fat or fluid present was small, with resultant issues in partial volume averaging.
| Conclusions | |
|---|
The diagnosis of malignancy is often made in a bone lesion subsequent to an initially indeterminate result at percutaneous CT-guided bone biopsy in a tertiary cancer center. This underscores the importance of vigorous pursuit of a definitive diagnosis after obtaining indeterminate percutaneous CT-guided bone biopsy results during the evaluation of a bone lesion. Absence of a cancer history, larger relative fraction of lesional sclerosis, and intralesional fat attenuation were strongly associated with indeterminate results, whereas cortical destruction and associated extraosseous mass were associated with diagnostic results. Increased age, female sex, and cancer history increased the likelihood of obtaining malignant results. A larger relative fraction of lesional sclerosis and a sclerotic rim, in contrast, were associated with a decreased likelihood of obtaining malignant results.
| References | |
|---|
