Diagnostic Accuracy and Clinical Utility of Biopsy in Musculoskeletal Lesions: A Comparison of Fine-Needle Aspiration, Core, and Open Biopsy Techniques Lester J. Layfield, M.D.,1 * Robert L. Schmidt, M.D., Nikhil Sangle, M.D.,2 and Julia R. Crim, M.D.3
Selection of biopsy technique for musculoskeletal lesions is complex. Fine-needle aspiration (FNA) is uncommonly used due to concerns regarding accuracy. We compared diagnostic accuracy of FNA, core, and open biopsy in a series of musculoskeletal lesions. Records of the University of Utah were searched for biopsy and resection specimens of musculoskeletal lesions. Results of corresponding imaging studies were obtained. Biopsy and FNA diagnoses were correlated with resection diagnoses. For each technique, diagnostic accuracy, utility, and frequency of subsequent biopsy were calculated. Open biopsy had the highest diagnostic accuracy (89%) followed by FNA (82%) and core biopsy (78%). Clinically significant errors occurred with all methods. The likelihood of an open biopsy being performed was affected by prior performance of an FNA or core biopsy and by diagnostic imaging and FNA results. Diagn. Cytopathol. 2014;42:476– 486. VC 2014 Wiley Periodicals, Inc. Key Words: studies
FNA; musculoskeletal; biopsy; accuracy; imaging
The selection of a biopsy technique for a musculoskeletal lesion is not a trivial decision and is based on a number of factors. The surgeon’s personal experience and the
1 Department of Pathology and Anatomical Sciences, University of Missouri, Columbia, Missouri 2 Department of Pathology, University of Utah School of Medicine and ARUP Laboratories, Salt Lake City, Utah 3 Department of Radiology, University of Missouri School of Medicine, Columbia, Missouri *Correspondence to: Lester J. Layfield, M.D., Professor and Chair, Department of Pathology and Anatomical Sciences, M263 Medical Science Building, One Hospital Drive, Columbia, MO 65212, USA. E-mail: [email protected] Disclosure: There are no disclaimers or conflicts of interest to report. Received 6 November 2012; Revised 1 March 2013; Accepted 4 April 2013 DOI: 10.1002/dc.23005 Published online 19 March 2014 in Wiley Online Library (wileyonlinelibrary.com).
C 2014 WILEY PERIODICALS, INC. V
2 Ph.D.,
pathologist’s familiarity with the cytologic and histopathologic appearance of musculoskeletal lesions are important factors in assessing which biopsy method will be used. Similarly, the location of the lesion, its proximity to vital structures, its imaging [magnetic resonance (MR) imaging] characteristics, potential for complications, and the perceived accuracies of the various biopsy methods all affect the selection of the biopsy technique. The optimal biopsy method would minimize the likelihood of complications while maximizing the probability of obtaining an accurate diagnosis. Ideally, a sequence of increasingly invasive procedures is used until a clinically accurate and useful diagnosis is obtained. Mankin et al.1 discussed the issues associated with biopsy of musculoskeletal lesions. Both core needle and open biopsy are associated with a significant error rate and complications which can alter the optimal treatment plan. In a study of 329 patients, 18.2% of biopsies were associated with significant diagnostic errors and an additional 10.3% had nonrepresentative or technically poor biopsies.1 Accuracy of biopsy result is an important consideration for selection of technique and the accuracy of FNA, core needle and open biopsy must be compared using similar endpoints. The current literature reports accuracies in several ways. Some studies report the accuracy of separating benign from malignant lesions while others report the accuracy in diagnosis of specific histologic types of neoplasms. Some studies include both primary and metastatic neoplasms while others only report data pertaining to primary musculoskeletal lesions. These differences in study design have significant impact on accuracy data. Accuracy can be defined as diagnostic accuracy (# correct=ðcorrect1 incorrectÞ) or therapeutic accuracy (ðcorrect1insignificant errorsÞ=ðcorrect1 incorrectÞ). Diagnostic Cytopathology, Vol. 42, No 6
476
Diagnostic Cytopathology DOI 10.1002/dc
DIAGNOSTIC ACCURACY AND CLINICAL UTILITY OF BIOPSY
Core needle biopsy has been demonstrated to be a highly accurate technique for the identification of musculoskeletal lesions.2–14 Reported accuracy rates have varied considerably depending on the definition of accuracy used in the report. Accuracy rates for separating benign from malignant lesions range up to the mid or upper 90s.4,5,11–14 Several authors have recommended core biopsy as an alternative to open biopsy or as a biopsy technique that should be used prior to open biopsy in most patients.3,6,7 Fine-needle aspiration (FNA) also is associated with good diagnostic accuracy.15–26 In some reports, diagnostic accuracy of FNA has been as high as 97%.21,25,26 Nonetheless, most studies directly comparing FNA and core needle biopsy have demonstrated a slight superiority for the core biopsy method.27–29 Clinical utility of a biopsy technique impacts whether a specific type of biopsy will be used by clinicians and determines the biopsy techniques’ value in the sequence of investigative studies. Ogilvie etal.30 documented that published series on diagnostic accuracy overestimate clinical utility of both FNA and core biopsy. From a clinician point of view, clinical utility of a test may be more important in a sequence of diagnostic examinations used to develop a treatment plan. Logan et al.31 have developed a biopsy algorithm using FNA and core needle biopsy. This biopsy algorithm has yielded an overall diagnostic accuracy of 96%. These authors have recommended using FNA for small soft tissue masses and core needle biopsy for larger masses over 3 cm. Other protocols exist using FNA, core biopsy and finally open biopsy in an ascending order of invasiveness and potential for patient morbidity. We investigated the test utility and diagnostic accuracy of imaging and biopsy alternatives to weigh diagnostic accuracy, therapeutic accuracy, and clinical utility of each technique in arriving at a clinically useful diagnosis. As discussed by Moons et al.,32 the majority of published studies describe test research rather than diagnostic research. “Test research” studies a single test or univariable approach, which focuses on a particular test to quantify its sensitivity, specificity, or diagnostic accuracy. This approach characterizes a test rather than estimating the test contribution to establishing a diagnosis. In this study, we are investigating “diagnostic contribution” of a number of tests and attempt to characterize a test’s contribution to the establishment of a clinically useful diagnosis. A diagnostic algorithm is also discussed.
Methods The records of the Department of Pathology of the University of Utah Hospital and Clinics were electronically searched for all FNAs, core and open biopsies and resection specimens of musculoskeletal lesions. The results of MR imaging studies were also obtained for all cases of
musculoskeletal lesions in which a biopsy or resection specimen was obtained. FNA and biopsy diagnoses were correlated with resection specimen diagnoses and the results of the other types of biopsies. For each technique, diagnostic accuracy, therapeutic accuracy, clinical utility, and frequency of subsequent additional biopsy were calculated. Diagnostic accuracy was defined as (# correct= ðcorrect1incorrectÞ) and therapeutic accuracy was calculated as ((correct 1 insignificant errors)/(correct 1 incorrect))). All incorrect diagnoses were classified as minor (clinically insignificant) or major (clinically significant). An error was judged as clinically significant if it altered optimal therapy or resulted in an additional biopsy. All other incorrect diagnoses, which did not impact patient outcome or therapy selected, were considered clinically insignificant. The impact of imaging studies on type of biopsy performed and whether or not additional biopsies were obtained was also recorded. The “clinical utility” of a test is defined as the percentage of cases in which the additional test was correct and, in addition, resulted in a change in therapy or resulted in a change from no diagnosis to a diagnosis. The “diagnostic utility” of a test is defined as the percentage of cases in which the additional test was correct and, in addition, resulted in a change of diagnosis (regardless of significance). A change from no diagnosis (noncontributory or indeterminate) to a specific diagnosis has diagnostic utility. Utility is defined incrementally with respect to a previous state of knowledge. Thus, if test 1 and test 2 are performed in sequence, one can only discuss the clinical utility of test 2. The clinical utility of test 1 with respect to test 2 is not defined. The clinical utility of test 1 can only be determined with respect to previous testing.
Statistical Methods The association between biopsy performance and predictor variables was determined using Pearson’s chi square test. All statistical tests were performed at the 0.05 significance level. Calculations were performed using Stata Release 12 (Stata Corp, College Station, TX). Data was maintained in Microsoft Access.
Cost-Accuracy Analysis Costs of FNA ($3490) and open biopsy ($5706) were based on published estimates.33 The cost of CNB was estimated at $4000 per biopsy. The therapeutic accuracy of combinations of tests were estimated by assuming that the tests were independent and calculating the joint distribution of outcomes from the marginal distributions of the individual tests. The therapeutic accuracy for a diagnostic pathway was estimated by repeating this procedure for each incremental test. The calculation of therapeutic accuracy for a combined test was determined by calculating the total probability of each outcome (therapeutically correct, therapeutically incorrect, and nondiagnostic). The Diagnostic Cytopathology, Vol. 42, No 6
477
Diagnostic Cytopathology DOI 10.1002/dc
LAYFIELD ET AL.
scoring of outcomes is shown in Table I. The calculation of outcome frequencies is illustrated in Table II. The overall accuracy of a sequence of tests was determined by repeating the calculations in Table II using each incremental test as “Test 2” and considering the combined accuracy of all previous tests as “Test 1.”
Results Patient Population Biopsy and resection specimen data were obtained from 257 patients. The mean patient age was 53.5 years (range 19–96). Figure 1 shows the age distribution for this study population. Figure 2 demonstrates the size distribution of the biopsied lesions as determined by MR. Two hundred fourteen of two hundred fifty seven cases (83%) had an MR performed. The majority (97%) of biopsied lesions were from the soft tissue. Because of the nature of the referral pattern to a musculoskeletal center, the majority of patients had sarcomas (76.7%) with 12.3% of lesions being of intermediate behavior and 11.0% benign. The single most common type of sarcoma was liposarcoma.
undergoing MR and those patients without MR (Fig. 3). Some patients had only FNA while others underwent FNA, core biopsy and open biopsy. Core biopsy was the single most common biopsy technique used. Table III shows the progression of diagnostic techniques used to obtain a final diagnosis. The majority of patients had an MR followed by either core needle biopsy or open biopsy to establish a diagnosis. When MR was not performed, open biopsy was the most frequent biopsy technique (Table III). FNA as a standalone diagnostic technique was rarely used (3 cases) but when coupled with MR examination was frequently used (Table III).
MR Results MR findings were characterized as either homogeneous (30% heterogeneous). Sixty-three percent (122 of 193) tumors were homogeneous and 37% were heterogeneous. Tissue type
Diagnostic Pathways The type and number of biopsy procedures varied from patient to patient and was different for those patients Table I. Scoring Scheme for Results of a Combined Test Test 2 Result
Test 1 Result
TC TI ND
TC
TI
ND
TC ND TC
ND TI TI
TC TI ND
The table shows the overall result when two tests are combined to form a single test. For example, the overall result for the combined result, (Test 1 5 TC; Test 2 5 TI), would be indeterminate. TC 5 clinically correct. TI 5 clinically incorrect. ND 5 nondiagnostic.
Fig. 1. Age distribution of patients included in the biopsy study population.
Table II. Relative Frequencies of Results for Combined Test Test 2 Combined test Test 1
TC TI ND
0.673 0.028 0.299
TC
TI
ND
0.710 0.478 0.020 0.212
0.161 0.108 0.005 0.048
0.129 0.087 0.004 0.039
TC 5 therapeutically correct. TI 5 therapeutically incorrect. ND 5 no diagnosis. In this example, the probability of obtaining a result of TC by Test 1 and a TC by Test 2 is 0.673 3 0.710 5 0.478. As shown in Table A, several different test combinations result in an overall result of TC (the underlined values in Table B correspond to the values that are therapeutically correct as shown in Table A). The overall probability that the combined test will produce a clinically correct result is found by summing the probabilities of each the ways that a result of TC can be obtained. In this example, Prob (TC) 5 0.478 1 0.087 1 0.212 5 0.777. Prob(ND) 5 0.108 1 0.020 1 0.039 5 0.167. Prob (TI) 5 0.005 1 0.004 1 0.048 5 0.056. Prob (TC) 1 Prob (TI) 1 Prob (ND) 5 1.
478
Diagnostic Cytopathology, Vol. 42, No 6
Fig. 2. Tumor size distribution in study as measured on MR images.
Diagnostic Cytopathology DOI 10.1002/dc
DIAGNOSTIC ACCURACY AND CLINICAL UTILITY OF BIOPSY
Fig. 3. The figure shows the distribution of procedures. Overlaps show cases having multiple procedures. For example, for those who had MR, 27 cases also had FNAs, 60 cases had an open biopsy, and 10 cases had both FNA and open biopsy.
Table III. Description of Diagnostic Pathways Path 1 2 3 4 5 6 7 8 9 10 11 12 13 14
MR x x x x x x x
FNA
CNB
Open
x x x
x x
x x x x
x x x x
x x
x x x x
x
x x x x
Frequency 85 27 17 60 12 11 2 13 3 1 25 0 1 0
Each line indicates a unique pattern of investigation.
diagnoses of 214 cases by MR were classified as correct (32.2%), incorrect (4.7%) and noncontributory (63.1%).
Accuracy Results Table IV compares the diagnostic accuracy of imaging studies and biopsy techniques. MR was noncontributory in 30% of cases, correct in 65% of cases, and incorrect in 4% of cases. Its overall diagnostic accuracy when
contributory was 94%. The associated therapeutic accuracy was 99%. Nine errors occurred in MR interpretation only one of which was significant. FNA was noncontributory in eight cases (13%), correct in 44 (71%) cases, and incorrect in 10 (16%) cases. Its overall diagnostic accuracy was 82%, its therapeutic accuracy 90% (Table IV). Core needle biopsy was noncontributory in a single case (0.8%), correct in 101 cases (78%), and incorrect in 28 cases (22%). Its diagnostic accuracy was 78% and therapeutic accuracy 85% (Table IV). In contrast, open biopsy contributed to the diagnosis in all cases. Open biopsy was correct in 99 cases (89%) and incorrect in 12 cases (11%). Open biopsy’s diagnostic accuracy was 89% and its therapeutic accuracy 91%. Significant errors occurred in five FNAs, 19 core needle biopsies and 10 open biopsies representing 8, 15, and 9% of procedures, respectively.
Characterization of Errors Tables V, VI, and VII list the diagnostic errors for FNA, core needle biopsy, and open biopsy with subsequent resection diagnosis. No biopsy technique was without error. FNA, core needle, and open biopsy all demonstrated both false positive and false negative diagnoses for malignancy. MR diagnostic errors most frequently Diagnostic Cytopathology, Vol. 42, No 6
479
Diagnostic Cytopathology DOI 10.1002/dc
480
Diagnostic accuracy 5 #correct/(#correct 1 #incorrect). Therapeutic accuracy 5 (#correct 1 #insignificant errors)/(#correct 1 #incorrect). Accuracy calculations are based on contributory cases. MR accuracy statistics are calculated for cases where MR was contributory.
18 (16.2%) 89 (80.2%) 0 35 (26.9%) 87 (66.9%) 2 (1.5%) 11 (5.1%) 134 (62.6%) 55(25.7%) Benign Malignant Noncontributory
Diagnostic Category Percent
b
10 2 4 (3.6%) 19 9 6 (4.6%) 5 5 13 (21.0%) 1 8 14 (6.6%) Error Significance
Diagnostic Cytopathology, Vol. 42, No 6
a
54 (21.0) 199 (77.4%) 0%
91.0% (101/111) [84–96%] 85.3% (110/129) [78–91%] 90.7% (49/54) [80–97%]
257 99.3% (148–149) [96–100%]
11 (17.7%) 31 (50.0%) 7 (11.3%)
4 (1.6%)
0 99 (89.2%) 12 (10.8%) 111 89.2% (99/111) [82–94%] 1 (0.8%) 101 (77.7%) 28 (21.5%) 130 78.3% (101/129) [70–85%] 8 (12.9%) 44 (71.0%) 10 (16.1%) 62 81.5% (44/54) [69–91%] 65 (30.4%) 140 (65.4%) 9 (4.2%) 214 94.0% (140/149) [89–97%]
Noncontributory Correct Incorrect Total cases Diagnosticb Accuracy [95% CI] Therapeuticb Accuracy [95% CI] Significant Not Significant Indeterminate Accuracy
Table IV. Comparison of Diagnostic Accuracy
MR
FNA
Resection Open CNB
LAYFIELD ET AL.
involved misinterpretation of a fatty neoplasm. FNA and core biopsy had similar types of errors and made errors on the same types of lesions. Table VIII summarizes the concordance data between methods. Concordance between MR and FNA was 89% while concordance between MR and core needle biopsy was 86%. The concordance between FNA and core needle biopsy was 73%.
Incremental Utility Tables IX and X summarize utility of open biopsy following FNA and open biopsy following core needle biopsy. Utility is defined incrementally with respect to the previous state of knowledge. Thus, when test 1 and test 2 are performed in sequence, one can determine the clinical utility of test 2 in relationship to the results of test 1. Open biopsy demonstrated diagnostic utility in 10 of 14 cases and clinical utility in 7 of 14 cases. As seen in Table X, open biopsy following core needle biopsy demonstrated diagnostic utility in 4 of 14 cases (29%), clinical utility in 3 of 14 cases (21%), and diagnostic and clinical disutility in 2 of 14 cases (7%).
Factors Affecting the Open Biopsy Rate A number of factors affected the probability that an open biopsy would be performed. These included the results of MR, FNA, and core needle biopsies. Table XI documents the factors affecting the probability of an open biopsy being performed. As would be expected, performance of a prior MR, FNA, or core needle biopsy all impacted whether or not an open biopsy would be performed to establish a final diagnosis. While diagnostic categories for FNA and core needle biopsy were not significant factors in affecting the probability that an open biopsy would be performed, the MR diagnostic category was. The MR diagnostic category open biopsy rate was associated with the MR diagnostic category (v2 511:5; p50:01Þ. When an MR result was characterized as benign or malignant, the likelihood of open biopsy being performed was reduced. When MR was characterized as noncontributory, a majority of patients underwent open biopsy. The open biopsy rate was also associated with the performance of MR v2 56:2; p50:01. When MR had been performed, only 40% (85/214) of patients underwent open biopsy. However, when MR was not performed, 60% (26/43) of patients underwent open biopsy. The performance of FNA had a similar impact on open biopsy rate (v2 514:1; p < 0:001). When patients underwent FNA of a lesion, only 16% (14/62) of patients required open biopsy but when no FNA was performed, half of all patients (97/195) underwent open biopsy. The open biopsy rate was also associated with the FNA diagnostic category and whether or not core biopsy was performed. When FNA category was either benign or malignant, open biopsy was unlikely but when FNA was categorized
Diagnostic Cytopathology DOI 10.1002/dc
DIAGNOSTIC ACCURACY AND CLINICAL UTILITY OF BIOPSY Table V. FNA Errors Significant?
CaseID
FNAdx
ResectionDx
10 59 66 67 94 156 196 198 246 270
Chondrosarcoma Ewing’s Schwannoma vs. low grade MPNST Spindle cell neoplasm Spindle cell neoplasm Benign nerve sheath tumor Inflammatory cyst Cyst with atypical cells High grade sarcoma Neuroendocrine neoplasm
Chondroblastic osteosarcoma Small round cell neoplasm Atypical myofibroblastic neoplasm Epithlioid sarcoma Myxoid liposarcoma with dedifferentiation Synovial sarcoma gr. 3 High grade sarcoma with focal myogenous diff. Hemangioma Diffuse large B cell lymphoma GIST
0 0 0 0 0 1 1 1 1 1
Table VI. Core Needle Biopsy Errors Significant?
CaseID
CNBdx
ResectionDx
No No No No No No No No No Yes Yes Yes Yes Yes Yes Yes Yes
24 63 82 88 126 139 210 218 275 1 18 19 47 92 94 99 136
MFH High grade myogenous sarcoma Malignant solitary fibrous tumor Malignant solitary fibrous tumor MFH Myxoid and vascular spindle cell neoplasm MPNST Solitary fibrous tumor No e/o neoplasm Extrabdominal fibromatosis, low grade SBRCT c/w Ewing’s Spindle cell sarcoma Lipoma Ewing’s/PNET Low grade myxoid spindle cell neoplasm Small round blue cell tumor Poorly diff. carcinoma
Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
153 158 224 236 244 246 248 260 266 267 276
Low grade chondromyxoid proliferation Lipoma Low grade cartilagenous neoplasm Low grade spindle cell proliferation Well differentiated liposarcoma High grade sarcoma No evidence of neoplasm Osteocartilagenous reparative process Neurofibroma No definite lesion Organizing hematoma
Pleomorphic rhabdomyosarcoma High grade MPNST Pleomorphic liposarcoma MPNST Leiomyosarcoma Solitary fibrous tumor Pleomorphic sarcoma with rhabdo. Features Ovarian fibroma Benign cystic lymphangioma Low grade unclassified spindle cell sarcoma Anaplastic large cell lymphoma Spindle cell melanoma Well diff. liposarcoma Leiomyosarcoma Myxoid liposarcoma with dedifferentiation Myxoid and round cell liposarcoma MPNST with rhabdomyoblastic differentiation (malignant triton tumor) Myxoid and round cell liposarcoma Well diff. liposarcoma Chondrosarcoma gr. 2 Dedifferentiated liposarcoma Well differentiated liposarcoma with focal dediff. Diffuse large B cell lymphoma Chordoma Extraskeletal osteosarcoma Neurofibroma progressed to MPNST Malignant poorly differentiated neoplasm Ossifying fibromyxoid tumor
Table VII. Open Biopsy Errors Significant?
CaseID
OpenBiopsyDx
ResectionDx High grade sarcoma with many areas of pleomorphic liposarcoma Malignant mesenchymoma of nerve sheath, high grade Aneurysmal fibrous histiocytoma High grade pleomorphic sarcoma Schwannoma with degeneration Round cell liposarcoma, high grade Melanoma Clear cell chondrosarcoma
0
43
Malignant solitary fibrous tumor
0 1 1 1 1 1 1
131 17 44 79 83 109 124
1
151
1 1 1
195 199 225
Low grade MPNST Ewing’s sarcoma/PNET Inflammatory pseudotumor MPNST Hemangiopericytoma Pleomorphic sarcoma gr. 3 Chondroblastoma; second Bx 5 atypical chondroblastic tumor Ewing’s sarcoma; second Bx 5 myxoid chondrosarcoma Low grade vascular neoplasm Chondroblastic osteosarcoma Well diff. liposarcoma
Myxoid chondrosarcoma Angiosarcoma Chondromyxoid fibroma Well diff. liposarcoma with focal dedifferentiation
Diagnostic Cytopathology, Vol. 42, No 6
481
Diagnostic Cytopathology DOI 10.1002/dc
LAYFIELD ET AL.
as indeterminate, nearly half of all patients underwent open biopsy. The impact of core needle biopsy on open biopsy rate was similar to that seen for FNA. When core needle biopsy was performed, only a minority of patients (11%) went on to open biopsy (v2 5112:7; p < 0:001Þ.
Cost-Accuracy Analysis of Diagnostic Pathways The estimated overall accuracy and cost of different diagnostic pathways are shown Figure 4. MR combined with FNA, CNB, or open biopsy provided the most efficient alternatives.
Discussion The work-up of a potentially neoplastic lesion within the musculoskeletal system is a complex process requiring Table VIII. Summary of Concordance statistics Method 1 Method 2 Discordant Concordant Noncontributory Total Concordance Rate (%)
MR FNA 4 33 21 58 89.2
MR CNB 13 83 22 118 86.4
FNA CNB 4 11 5 20 73.3
input from a team of surgeons, radiologists, and pathologists. A variety of techniques are available for the workup of musculoskeletal lesions including imaging techniques (MR imaging, computerized tomography and plain films), FNA, core needle biopsy and open biopsy. The selection of the appropriate technique or techniques depends upon the surgeon’s personal experience and the knowledge/skill level of the assisting radiologist and pathologist. Optimally, the combination of techniques used should maximize the potential for high diagnostic accuracy while minimizing cost, potential for technical complications and the invasiveness of the procedure selected. In most cases, a sequence of diagnostic procedures will be used running from least invasive (imaging studies) to most invasive (open biopsy). Each step in this sequence of procedures will influence whether or not a more invasive technique will be ordered to establish a diagnosis. In general, the sequence of these diagnostic techniques will be MR, FNA, core needle biopsy and finally, open biopsy. Table III describes the diagnostic pathways used for the work-up of patients in this study. In our study, most patients underwent MR followed by core needle
Table IX. Summary of Utility of Open Biopsy following FNA
Case ID 4 16 27 29 34
FNAdx Spindle cell sarcoma Mesenchymal neoplasm Fibrous tissue
FNAdx correct
FNAdx significant Open biopsy error diagnosis
Yes Yes Noncontributory
Resection diagnosis
Synovial sarcoma Synovial sarcoma
Yes Yes
Synovial sarcoma Synovial sarcoma
Leiomyosarcoma gr. 2 Myxoid liposarcoma Ewing’s sarcoma
Yes
Low grade fibromyxoid sarcoma gr. 1 Pleomorphic sarcoma gr. 3 Desmoplastic small round cell tumor Atypical myofibroblastic neoplasm
Yes
Yes
Yes
Leiomyosarcoma gr. 3 Myxoid liposarcoma No residual tumor
Yes
Yes
Low grade fibromyxoid sarcoma, gr. 1 Pleomorphic sarcoma gr. 3 Small round cell neoplasm Atypical myofibroblastic neoplasm
Yes
Yes
37
Noncontributory
57
Pleomorphic sarcoma
Yes
59
Ewing’s
No
66
No No
Epithelioid sarcoma
Yes
Epithlioid sarcoma
Yes
76
Schwannoma vs. low No grade MPNST Spindle cell No neoplasm Atypical cells present Noncontributory
Yes
Yes
Inflammatory cyst
No
Yes
Yes
Yes
205
Rare atypical cells
Noncontributory
Amyloid with tissue reaction
Yes
Yes
Yes
249
Necrotic neoplasm
Yes
Malignant neoplasm with neuroendocrine features
Yes
Dedifferentiated liposarcoma High grade sarcoma with focal myogenous diff. Amyloidoma with plasma cell dyscrasia Ewings sarcoma/ PNET
Yes
196
Dedifferentiated liposarcoma High grade sarcoma
Yes
Yes
No
Summary: 10 of 14 open biopsies following FNA had diagnostic utility. 7 of 14 open biopsies following FNA had clinical utility.
482
Diagnostic Cytopathology, Vol. 42, No 6
Yes Yes
Diagnostic Clinical utility? utility?
Myxoid liposarcoma Small blue cell neoplasm Insufficient
67
Yes Yes
OB OB is significant correct error
Yes Yes Yes
Yes
Yes Yes
Diagnostic Cytopathology DOI 10.1002/dc
??? ???
Yes Yes
4 of 14 open biopsies following CNB had diagnostic utility. 3 of 14 open biopsies following CNB had clinical utility. 2 of 14 open biopsies following CNB had diagnostic and clinical disutility.
Yes No Yes Yes Myxoid liposarcoma Cartilage forming neoplasm No e/o neoplasm Malignant poorly diff. neoplasm 160 199 205 274
Yes Yes Yes Yes
Yes Leiomyosarcoma Poorly diff. carcinoma 101 136
Yes No
No Yes
Myxoid and round cell liposarcoma Chondroblastic osteosarcoma Amyloid with tissue reaction Necrosis, rare malignant cells
Yes Yes
Yes
Yes
Leiomyosarcoma gr. 3 MPNST with rhabdomyoblastic differentiation (malignant triton tumor) Myxoid liposarcoma Chondromyxoid fibroma Amyloidoma with plasma cell Dyscrasia Angiosarcoma
Yes Yes Yes
??? ???
Yes Yes
Myxoma Anaplastic large cell lymphoma MPNST High grade pleomorphic sarcoma Synovial sarcoma Dedifferentiated liposarcoma Pleomorphic liposarcoma Myxoid and round cell liposarcoma Yes Yes 1 Yes No Yes Yes Yes Yes
Myxoma Anaplastic large cell lymphoma MPNST Inflammatory pseudotumor Monophasic synovial sarcoma Dedifferentiated liposarcoma High grade sarcoma Myxoid liposarcoma vs. chondrosarcoma Leiomyosarcoma gr. 3 High grade sarcoma Yes
Yes No Yes Yes Yes Yes No No Myxoid neoplasm SBRCT c/w Ewing’s MPNST Spindle cell sarcoma gr.2/3 High grade sarcoma Pleomorphic spindle cell neoplasm Malignant solitary fibrous tumor Small round blue cell tumor
CNBdx Case ID
11 18 25 44 72 76 82 99
CNB Signif Error CNBdx Correct
Table X. Summary of Utility of Open Biopsy following CNB
Open Biopsy Dx
OB dx Is Correct
OB Dx Signif Error
Resection Dx
Diagnostic Utility
Clinical Utility
DIAGNOSTIC ACCURACY AND CLINICAL UTILITY OF BIOPSY
biopsy. The second most frequent pathway was MR followed by open biopsy. MR followed by FNA was the third most frequent diagnostic pathway selected. Figure 3 illustrates the relationship of the various diagnostic procedures with and without the use of MR. The selection of biopsy technique for a musculoskeletal lesion is not a trivial issue and Mankin et al.1 documented the hazards of both core needle biopsy and open biopsy. Both techniques are associated with significant error and complication rates. Many of these caused the optimal treatment plan to be modified. In his multi-institutional study of 329 patients, fully 18% of biopsies were associated with a significant diagnostic error and an additional 10% resulted in a non-representative or technically poor biopsy.1 Mankin also reported that a significant number of patients undergoing core needle or open biopsy had procedural complications, which subsequently altered the treatment plan. Accuracy of a diagnostic technique can be variably defined as diagnostic accuracy or therapeutic accuracy. Diagnostic accuracy specifically documents the number of correct cases as a percentage of total cases. Therapeutic accuracy is clinically more important in that it reports the number of correct cases and insignificant errors as a percentage of total cases studied. Accuracy statistics for core needle biopsy show it to be a highly reliable technique for the identification of musculoskeletal lesions.2–14 Nonetheless, accuracy rates for separating benign from malignant lesions have varied considerably but most studies report accuracies above 90%.4,5,11–14 Based on these statistics, a number of authors have recommended core needle biopsy as an alternative to open biopsy.3,6,7 In many institutions including our own, core needle biopsy is performed on the majority of patients and is frequently the definitive biopsy technique used. More recently, FNA has been used in the diagnostic pathway for the investigation of musculoskeletal lesions. FNA is associated with good diagnostic accuracy15–26 and in some reports, has a diagnostic accuracy above 90%.21,25,26 However, some surgeons remain reticent to use FNA as the definitive diagnostic technique. Most studies directly comparing FNA and core needle biopsy have demonstrated a slight superiority in diagnostic accuracy for core needle biopsy.27–29 Logan et al.30 developed a biopsy algorithm using FNA and core needle biopsy. Logan’s algorithm yields an overall diagnostic accuracy of 96%.30 Our study included 257 patients referred to a musculoskeletal center. The majority of patients had soft tissue lesions (97%) and most were malignant (77%). The study protocol required a resection specimen for inclusion and this skewed the study population toward malignancies. At the time of the study, no protocol existed to determine which biopsy techniques were used and in what order. The
Diagnostic Cytopathology, Vol. 42, No 6
483
Diagnostic Cytopathology DOI 10.1002/dc
LAYFIELD ET AL. Table XI. Factors Affecting the Probability of an Open Biopsy Factor FNA CNB concordance MR CNB concordance MR FNA concordance Whether MR was performed Whether FNA was performed Whether CNB was performed MR Diagnostic category CNB Diagnostic category FNA Diagnostic category
Conclusion
P value
Not significant Not significant Not significant Significant Significant Significant Significant Not significant Not significant
0.25 0.39 0.18 0.012
Selection of biopsy technique for musculoskeletal lesions is complex. Fine-needle aspiration (FNA) is uncommonly used due to concerns regarding accuracy. We compared diagnostic accuracy of FNA, core, and open biopsy in a series of musculoskeletal lesions. Records of the University of Utah were searched for biopsy and resection specimens of musculoskeletal lesions. Results of corresponding imaging studies were obtained. Biopsy and FNA diagnoses were correlated with resection diagnoses. For each technique, diagnostic accuracy, utility, and frequency of subsequent biopsy were calculated. Open biopsy had the highest diagnostic accuracy (89%) followed by FNA (82%) and core biopsy (78%). Clinically significant errors occurred with all methods. The likelihood of an open biopsy being performed was affected by prior performance of an FNA or core biopsy and by diagnostic imaging and FNA results. Diagn. Cytopathol. 2014;42:476– 486. VC 2014 Wiley Periodicals, Inc. Key Words: studies
FNA; musculoskeletal; biopsy; accuracy; imaging
The selection of a biopsy technique for a musculoskeletal lesion is not a trivial decision and is based on a number of factors. The surgeon’s personal experience and the
1 Department of Pathology and Anatomical Sciences, University of Missouri, Columbia, Missouri 2 Department of Pathology, University of Utah School of Medicine and ARUP Laboratories, Salt Lake City, Utah 3 Department of Radiology, University of Missouri School of Medicine, Columbia, Missouri *Correspondence to: Lester J. Layfield, M.D., Professor and Chair, Department of Pathology and Anatomical Sciences, M263 Medical Science Building, One Hospital Drive, Columbia, MO 65212, USA. E-mail: [email protected] Disclosure: There are no disclaimers or conflicts of interest to report. Received 6 November 2012; Revised 1 March 2013; Accepted 4 April 2013 DOI: 10.1002/dc.23005 Published online 19 March 2014 in Wiley Online Library (wileyonlinelibrary.com).
C 2014 WILEY PERIODICALS, INC. V
2 Ph.D.,
pathologist’s familiarity with the cytologic and histopathologic appearance of musculoskeletal lesions are important factors in assessing which biopsy method will be used. Similarly, the location of the lesion, its proximity to vital structures, its imaging [magnetic resonance (MR) imaging] characteristics, potential for complications, and the perceived accuracies of the various biopsy methods all affect the selection of the biopsy technique. The optimal biopsy method would minimize the likelihood of complications while maximizing the probability of obtaining an accurate diagnosis. Ideally, a sequence of increasingly invasive procedures is used until a clinically accurate and useful diagnosis is obtained. Mankin et al.1 discussed the issues associated with biopsy of musculoskeletal lesions. Both core needle and open biopsy are associated with a significant error rate and complications which can alter the optimal treatment plan. In a study of 329 patients, 18.2% of biopsies were associated with significant diagnostic errors and an additional 10.3% had nonrepresentative or technically poor biopsies.1 Accuracy of biopsy result is an important consideration for selection of technique and the accuracy of FNA, core needle and open biopsy must be compared using similar endpoints. The current literature reports accuracies in several ways. Some studies report the accuracy of separating benign from malignant lesions while others report the accuracy in diagnosis of specific histologic types of neoplasms. Some studies include both primary and metastatic neoplasms while others only report data pertaining to primary musculoskeletal lesions. These differences in study design have significant impact on accuracy data. Accuracy can be defined as diagnostic accuracy (# correct=ðcorrect1 incorrectÞ) or therapeutic accuracy (ðcorrect1insignificant errorsÞ=ðcorrect1 incorrectÞ). Diagnostic Cytopathology, Vol. 42, No 6
476
Diagnostic Cytopathology DOI 10.1002/dc
DIAGNOSTIC ACCURACY AND CLINICAL UTILITY OF BIOPSY
Core needle biopsy has been demonstrated to be a highly accurate technique for the identification of musculoskeletal lesions.2–14 Reported accuracy rates have varied considerably depending on the definition of accuracy used in the report. Accuracy rates for separating benign from malignant lesions range up to the mid or upper 90s.4,5,11–14 Several authors have recommended core biopsy as an alternative to open biopsy or as a biopsy technique that should be used prior to open biopsy in most patients.3,6,7 Fine-needle aspiration (FNA) also is associated with good diagnostic accuracy.15–26 In some reports, diagnostic accuracy of FNA has been as high as 97%.21,25,26 Nonetheless, most studies directly comparing FNA and core needle biopsy have demonstrated a slight superiority for the core biopsy method.27–29 Clinical utility of a biopsy technique impacts whether a specific type of biopsy will be used by clinicians and determines the biopsy techniques’ value in the sequence of investigative studies. Ogilvie etal.30 documented that published series on diagnostic accuracy overestimate clinical utility of both FNA and core biopsy. From a clinician point of view, clinical utility of a test may be more important in a sequence of diagnostic examinations used to develop a treatment plan. Logan et al.31 have developed a biopsy algorithm using FNA and core needle biopsy. This biopsy algorithm has yielded an overall diagnostic accuracy of 96%. These authors have recommended using FNA for small soft tissue masses and core needle biopsy for larger masses over 3 cm. Other protocols exist using FNA, core biopsy and finally open biopsy in an ascending order of invasiveness and potential for patient morbidity. We investigated the test utility and diagnostic accuracy of imaging and biopsy alternatives to weigh diagnostic accuracy, therapeutic accuracy, and clinical utility of each technique in arriving at a clinically useful diagnosis. As discussed by Moons et al.,32 the majority of published studies describe test research rather than diagnostic research. “Test research” studies a single test or univariable approach, which focuses on a particular test to quantify its sensitivity, specificity, or diagnostic accuracy. This approach characterizes a test rather than estimating the test contribution to establishing a diagnosis. In this study, we are investigating “diagnostic contribution” of a number of tests and attempt to characterize a test’s contribution to the establishment of a clinically useful diagnosis. A diagnostic algorithm is also discussed.
Methods The records of the Department of Pathology of the University of Utah Hospital and Clinics were electronically searched for all FNAs, core and open biopsies and resection specimens of musculoskeletal lesions. The results of MR imaging studies were also obtained for all cases of
musculoskeletal lesions in which a biopsy or resection specimen was obtained. FNA and biopsy diagnoses were correlated with resection specimen diagnoses and the results of the other types of biopsies. For each technique, diagnostic accuracy, therapeutic accuracy, clinical utility, and frequency of subsequent additional biopsy were calculated. Diagnostic accuracy was defined as (# correct= ðcorrect1incorrectÞ) and therapeutic accuracy was calculated as ((correct 1 insignificant errors)/(correct 1 incorrect))). All incorrect diagnoses were classified as minor (clinically insignificant) or major (clinically significant). An error was judged as clinically significant if it altered optimal therapy or resulted in an additional biopsy. All other incorrect diagnoses, which did not impact patient outcome or therapy selected, were considered clinically insignificant. The impact of imaging studies on type of biopsy performed and whether or not additional biopsies were obtained was also recorded. The “clinical utility” of a test is defined as the percentage of cases in which the additional test was correct and, in addition, resulted in a change in therapy or resulted in a change from no diagnosis to a diagnosis. The “diagnostic utility” of a test is defined as the percentage of cases in which the additional test was correct and, in addition, resulted in a change of diagnosis (regardless of significance). A change from no diagnosis (noncontributory or indeterminate) to a specific diagnosis has diagnostic utility. Utility is defined incrementally with respect to a previous state of knowledge. Thus, if test 1 and test 2 are performed in sequence, one can only discuss the clinical utility of test 2. The clinical utility of test 1 with respect to test 2 is not defined. The clinical utility of test 1 can only be determined with respect to previous testing.
Statistical Methods The association between biopsy performance and predictor variables was determined using Pearson’s chi square test. All statistical tests were performed at the 0.05 significance level. Calculations were performed using Stata Release 12 (Stata Corp, College Station, TX). Data was maintained in Microsoft Access.
Cost-Accuracy Analysis Costs of FNA ($3490) and open biopsy ($5706) were based on published estimates.33 The cost of CNB was estimated at $4000 per biopsy. The therapeutic accuracy of combinations of tests were estimated by assuming that the tests were independent and calculating the joint distribution of outcomes from the marginal distributions of the individual tests. The therapeutic accuracy for a diagnostic pathway was estimated by repeating this procedure for each incremental test. The calculation of therapeutic accuracy for a combined test was determined by calculating the total probability of each outcome (therapeutically correct, therapeutically incorrect, and nondiagnostic). The Diagnostic Cytopathology, Vol. 42, No 6
477
Diagnostic Cytopathology DOI 10.1002/dc
LAYFIELD ET AL.
scoring of outcomes is shown in Table I. The calculation of outcome frequencies is illustrated in Table II. The overall accuracy of a sequence of tests was determined by repeating the calculations in Table II using each incremental test as “Test 2” and considering the combined accuracy of all previous tests as “Test 1.”
Results Patient Population Biopsy and resection specimen data were obtained from 257 patients. The mean patient age was 53.5 years (range 19–96). Figure 1 shows the age distribution for this study population. Figure 2 demonstrates the size distribution of the biopsied lesions as determined by MR. Two hundred fourteen of two hundred fifty seven cases (83%) had an MR performed. The majority (97%) of biopsied lesions were from the soft tissue. Because of the nature of the referral pattern to a musculoskeletal center, the majority of patients had sarcomas (76.7%) with 12.3% of lesions being of intermediate behavior and 11.0% benign. The single most common type of sarcoma was liposarcoma.
undergoing MR and those patients without MR (Fig. 3). Some patients had only FNA while others underwent FNA, core biopsy and open biopsy. Core biopsy was the single most common biopsy technique used. Table III shows the progression of diagnostic techniques used to obtain a final diagnosis. The majority of patients had an MR followed by either core needle biopsy or open biopsy to establish a diagnosis. When MR was not performed, open biopsy was the most frequent biopsy technique (Table III). FNA as a standalone diagnostic technique was rarely used (3 cases) but when coupled with MR examination was frequently used (Table III).
MR Results MR findings were characterized as either homogeneous (30% heterogeneous). Sixty-three percent (122 of 193) tumors were homogeneous and 37% were heterogeneous. Tissue type
Diagnostic Pathways The type and number of biopsy procedures varied from patient to patient and was different for those patients Table I. Scoring Scheme for Results of a Combined Test Test 2 Result
Test 1 Result
TC TI ND
TC
TI
ND
TC ND TC
ND TI TI
TC TI ND
The table shows the overall result when two tests are combined to form a single test. For example, the overall result for the combined result, (Test 1 5 TC; Test 2 5 TI), would be indeterminate. TC 5 clinically correct. TI 5 clinically incorrect. ND 5 nondiagnostic.
Fig. 1. Age distribution of patients included in the biopsy study population.
Table II. Relative Frequencies of Results for Combined Test Test 2 Combined test Test 1
TC TI ND
0.673 0.028 0.299
TC
TI
ND
0.710 0.478 0.020 0.212
0.161 0.108 0.005 0.048
0.129 0.087 0.004 0.039
TC 5 therapeutically correct. TI 5 therapeutically incorrect. ND 5 no diagnosis. In this example, the probability of obtaining a result of TC by Test 1 and a TC by Test 2 is 0.673 3 0.710 5 0.478. As shown in Table A, several different test combinations result in an overall result of TC (the underlined values in Table B correspond to the values that are therapeutically correct as shown in Table A). The overall probability that the combined test will produce a clinically correct result is found by summing the probabilities of each the ways that a result of TC can be obtained. In this example, Prob (TC) 5 0.478 1 0.087 1 0.212 5 0.777. Prob(ND) 5 0.108 1 0.020 1 0.039 5 0.167. Prob (TI) 5 0.005 1 0.004 1 0.048 5 0.056. Prob (TC) 1 Prob (TI) 1 Prob (ND) 5 1.
478
Diagnostic Cytopathology, Vol. 42, No 6
Fig. 2. Tumor size distribution in study as measured on MR images.
Diagnostic Cytopathology DOI 10.1002/dc
DIAGNOSTIC ACCURACY AND CLINICAL UTILITY OF BIOPSY
Fig. 3. The figure shows the distribution of procedures. Overlaps show cases having multiple procedures. For example, for those who had MR, 27 cases also had FNAs, 60 cases had an open biopsy, and 10 cases had both FNA and open biopsy.
Table III. Description of Diagnostic Pathways Path 1 2 3 4 5 6 7 8 9 10 11 12 13 14
MR x x x x x x x
FNA
CNB
Open
x x x
x x
x x x x
x x x x
x x
x x x x
x
x x x x
Frequency 85 27 17 60 12 11 2 13 3 1 25 0 1 0
Each line indicates a unique pattern of investigation.
diagnoses of 214 cases by MR were classified as correct (32.2%), incorrect (4.7%) and noncontributory (63.1%).
Accuracy Results Table IV compares the diagnostic accuracy of imaging studies and biopsy techniques. MR was noncontributory in 30% of cases, correct in 65% of cases, and incorrect in 4% of cases. Its overall diagnostic accuracy when
contributory was 94%. The associated therapeutic accuracy was 99%. Nine errors occurred in MR interpretation only one of which was significant. FNA was noncontributory in eight cases (13%), correct in 44 (71%) cases, and incorrect in 10 (16%) cases. Its overall diagnostic accuracy was 82%, its therapeutic accuracy 90% (Table IV). Core needle biopsy was noncontributory in a single case (0.8%), correct in 101 cases (78%), and incorrect in 28 cases (22%). Its diagnostic accuracy was 78% and therapeutic accuracy 85% (Table IV). In contrast, open biopsy contributed to the diagnosis in all cases. Open biopsy was correct in 99 cases (89%) and incorrect in 12 cases (11%). Open biopsy’s diagnostic accuracy was 89% and its therapeutic accuracy 91%. Significant errors occurred in five FNAs, 19 core needle biopsies and 10 open biopsies representing 8, 15, and 9% of procedures, respectively.
Characterization of Errors Tables V, VI, and VII list the diagnostic errors for FNA, core needle biopsy, and open biopsy with subsequent resection diagnosis. No biopsy technique was without error. FNA, core needle, and open biopsy all demonstrated both false positive and false negative diagnoses for malignancy. MR diagnostic errors most frequently Diagnostic Cytopathology, Vol. 42, No 6
479
Diagnostic Cytopathology DOI 10.1002/dc
480
Diagnostic accuracy 5 #correct/(#correct 1 #incorrect). Therapeutic accuracy 5 (#correct 1 #insignificant errors)/(#correct 1 #incorrect). Accuracy calculations are based on contributory cases. MR accuracy statistics are calculated for cases where MR was contributory.
18 (16.2%) 89 (80.2%) 0 35 (26.9%) 87 (66.9%) 2 (1.5%) 11 (5.1%) 134 (62.6%) 55(25.7%) Benign Malignant Noncontributory
Diagnostic Category Percent
b
10 2 4 (3.6%) 19 9 6 (4.6%) 5 5 13 (21.0%) 1 8 14 (6.6%) Error Significance
Diagnostic Cytopathology, Vol. 42, No 6
a
54 (21.0) 199 (77.4%) 0%
91.0% (101/111) [84–96%] 85.3% (110/129) [78–91%] 90.7% (49/54) [80–97%]
257 99.3% (148–149) [96–100%]
11 (17.7%) 31 (50.0%) 7 (11.3%)
4 (1.6%)
0 99 (89.2%) 12 (10.8%) 111 89.2% (99/111) [82–94%] 1 (0.8%) 101 (77.7%) 28 (21.5%) 130 78.3% (101/129) [70–85%] 8 (12.9%) 44 (71.0%) 10 (16.1%) 62 81.5% (44/54) [69–91%] 65 (30.4%) 140 (65.4%) 9 (4.2%) 214 94.0% (140/149) [89–97%]
Noncontributory Correct Incorrect Total cases Diagnosticb Accuracy [95% CI] Therapeuticb Accuracy [95% CI] Significant Not Significant Indeterminate Accuracy
Table IV. Comparison of Diagnostic Accuracy
MR
FNA
Resection Open CNB
LAYFIELD ET AL.
involved misinterpretation of a fatty neoplasm. FNA and core biopsy had similar types of errors and made errors on the same types of lesions. Table VIII summarizes the concordance data between methods. Concordance between MR and FNA was 89% while concordance between MR and core needle biopsy was 86%. The concordance between FNA and core needle biopsy was 73%.
Incremental Utility Tables IX and X summarize utility of open biopsy following FNA and open biopsy following core needle biopsy. Utility is defined incrementally with respect to the previous state of knowledge. Thus, when test 1 and test 2 are performed in sequence, one can determine the clinical utility of test 2 in relationship to the results of test 1. Open biopsy demonstrated diagnostic utility in 10 of 14 cases and clinical utility in 7 of 14 cases. As seen in Table X, open biopsy following core needle biopsy demonstrated diagnostic utility in 4 of 14 cases (29%), clinical utility in 3 of 14 cases (21%), and diagnostic and clinical disutility in 2 of 14 cases (7%).
Factors Affecting the Open Biopsy Rate A number of factors affected the probability that an open biopsy would be performed. These included the results of MR, FNA, and core needle biopsies. Table XI documents the factors affecting the probability of an open biopsy being performed. As would be expected, performance of a prior MR, FNA, or core needle biopsy all impacted whether or not an open biopsy would be performed to establish a final diagnosis. While diagnostic categories for FNA and core needle biopsy were not significant factors in affecting the probability that an open biopsy would be performed, the MR diagnostic category was. The MR diagnostic category open biopsy rate was associated with the MR diagnostic category (v2 511:5; p50:01Þ. When an MR result was characterized as benign or malignant, the likelihood of open biopsy being performed was reduced. When MR was characterized as noncontributory, a majority of patients underwent open biopsy. The open biopsy rate was also associated with the performance of MR v2 56:2; p50:01. When MR had been performed, only 40% (85/214) of patients underwent open biopsy. However, when MR was not performed, 60% (26/43) of patients underwent open biopsy. The performance of FNA had a similar impact on open biopsy rate (v2 514:1; p < 0:001). When patients underwent FNA of a lesion, only 16% (14/62) of patients required open biopsy but when no FNA was performed, half of all patients (97/195) underwent open biopsy. The open biopsy rate was also associated with the FNA diagnostic category and whether or not core biopsy was performed. When FNA category was either benign or malignant, open biopsy was unlikely but when FNA was categorized
Diagnostic Cytopathology DOI 10.1002/dc
DIAGNOSTIC ACCURACY AND CLINICAL UTILITY OF BIOPSY Table V. FNA Errors Significant?
CaseID
FNAdx
ResectionDx
10 59 66 67 94 156 196 198 246 270
Chondrosarcoma Ewing’s Schwannoma vs. low grade MPNST Spindle cell neoplasm Spindle cell neoplasm Benign nerve sheath tumor Inflammatory cyst Cyst with atypical cells High grade sarcoma Neuroendocrine neoplasm
Chondroblastic osteosarcoma Small round cell neoplasm Atypical myofibroblastic neoplasm Epithlioid sarcoma Myxoid liposarcoma with dedifferentiation Synovial sarcoma gr. 3 High grade sarcoma with focal myogenous diff. Hemangioma Diffuse large B cell lymphoma GIST
0 0 0 0 0 1 1 1 1 1
Table VI. Core Needle Biopsy Errors Significant?
CaseID
CNBdx
ResectionDx
No No No No No No No No No Yes Yes Yes Yes Yes Yes Yes Yes
24 63 82 88 126 139 210 218 275 1 18 19 47 92 94 99 136
MFH High grade myogenous sarcoma Malignant solitary fibrous tumor Malignant solitary fibrous tumor MFH Myxoid and vascular spindle cell neoplasm MPNST Solitary fibrous tumor No e/o neoplasm Extrabdominal fibromatosis, low grade SBRCT c/w Ewing’s Spindle cell sarcoma Lipoma Ewing’s/PNET Low grade myxoid spindle cell neoplasm Small round blue cell tumor Poorly diff. carcinoma
Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
153 158 224 236 244 246 248 260 266 267 276
Low grade chondromyxoid proliferation Lipoma Low grade cartilagenous neoplasm Low grade spindle cell proliferation Well differentiated liposarcoma High grade sarcoma No evidence of neoplasm Osteocartilagenous reparative process Neurofibroma No definite lesion Organizing hematoma
Pleomorphic rhabdomyosarcoma High grade MPNST Pleomorphic liposarcoma MPNST Leiomyosarcoma Solitary fibrous tumor Pleomorphic sarcoma with rhabdo. Features Ovarian fibroma Benign cystic lymphangioma Low grade unclassified spindle cell sarcoma Anaplastic large cell lymphoma Spindle cell melanoma Well diff. liposarcoma Leiomyosarcoma Myxoid liposarcoma with dedifferentiation Myxoid and round cell liposarcoma MPNST with rhabdomyoblastic differentiation (malignant triton tumor) Myxoid and round cell liposarcoma Well diff. liposarcoma Chondrosarcoma gr. 2 Dedifferentiated liposarcoma Well differentiated liposarcoma with focal dediff. Diffuse large B cell lymphoma Chordoma Extraskeletal osteosarcoma Neurofibroma progressed to MPNST Malignant poorly differentiated neoplasm Ossifying fibromyxoid tumor
Table VII. Open Biopsy Errors Significant?
CaseID
OpenBiopsyDx
ResectionDx High grade sarcoma with many areas of pleomorphic liposarcoma Malignant mesenchymoma of nerve sheath, high grade Aneurysmal fibrous histiocytoma High grade pleomorphic sarcoma Schwannoma with degeneration Round cell liposarcoma, high grade Melanoma Clear cell chondrosarcoma
0
43
Malignant solitary fibrous tumor
0 1 1 1 1 1 1
131 17 44 79 83 109 124
1
151
1 1 1
195 199 225
Low grade MPNST Ewing’s sarcoma/PNET Inflammatory pseudotumor MPNST Hemangiopericytoma Pleomorphic sarcoma gr. 3 Chondroblastoma; second Bx 5 atypical chondroblastic tumor Ewing’s sarcoma; second Bx 5 myxoid chondrosarcoma Low grade vascular neoplasm Chondroblastic osteosarcoma Well diff. liposarcoma
Myxoid chondrosarcoma Angiosarcoma Chondromyxoid fibroma Well diff. liposarcoma with focal dedifferentiation
Diagnostic Cytopathology, Vol. 42, No 6
481
Diagnostic Cytopathology DOI 10.1002/dc
LAYFIELD ET AL.
as indeterminate, nearly half of all patients underwent open biopsy. The impact of core needle biopsy on open biopsy rate was similar to that seen for FNA. When core needle biopsy was performed, only a minority of patients (11%) went on to open biopsy (v2 5112:7; p < 0:001Þ.
Cost-Accuracy Analysis of Diagnostic Pathways The estimated overall accuracy and cost of different diagnostic pathways are shown Figure 4. MR combined with FNA, CNB, or open biopsy provided the most efficient alternatives.
Discussion The work-up of a potentially neoplastic lesion within the musculoskeletal system is a complex process requiring Table VIII. Summary of Concordance statistics Method 1 Method 2 Discordant Concordant Noncontributory Total Concordance Rate (%)
MR FNA 4 33 21 58 89.2
MR CNB 13 83 22 118 86.4
FNA CNB 4 11 5 20 73.3
input from a team of surgeons, radiologists, and pathologists. A variety of techniques are available for the workup of musculoskeletal lesions including imaging techniques (MR imaging, computerized tomography and plain films), FNA, core needle biopsy and open biopsy. The selection of the appropriate technique or techniques depends upon the surgeon’s personal experience and the knowledge/skill level of the assisting radiologist and pathologist. Optimally, the combination of techniques used should maximize the potential for high diagnostic accuracy while minimizing cost, potential for technical complications and the invasiveness of the procedure selected. In most cases, a sequence of diagnostic procedures will be used running from least invasive (imaging studies) to most invasive (open biopsy). Each step in this sequence of procedures will influence whether or not a more invasive technique will be ordered to establish a diagnosis. In general, the sequence of these diagnostic techniques will be MR, FNA, core needle biopsy and finally, open biopsy. Table III describes the diagnostic pathways used for the work-up of patients in this study. In our study, most patients underwent MR followed by core needle
Table IX. Summary of Utility of Open Biopsy following FNA
Case ID 4 16 27 29 34
FNAdx Spindle cell sarcoma Mesenchymal neoplasm Fibrous tissue
FNAdx correct
FNAdx significant Open biopsy error diagnosis
Yes Yes Noncontributory
Resection diagnosis
Synovial sarcoma Synovial sarcoma
Yes Yes
Synovial sarcoma Synovial sarcoma
Leiomyosarcoma gr. 2 Myxoid liposarcoma Ewing’s sarcoma
Yes
Low grade fibromyxoid sarcoma gr. 1 Pleomorphic sarcoma gr. 3 Desmoplastic small round cell tumor Atypical myofibroblastic neoplasm
Yes
Yes
Yes
Leiomyosarcoma gr. 3 Myxoid liposarcoma No residual tumor
Yes
Yes
Low grade fibromyxoid sarcoma, gr. 1 Pleomorphic sarcoma gr. 3 Small round cell neoplasm Atypical myofibroblastic neoplasm
Yes
Yes
37
Noncontributory
57
Pleomorphic sarcoma
Yes
59
Ewing’s
No
66
No No
Epithelioid sarcoma
Yes
Epithlioid sarcoma
Yes
76
Schwannoma vs. low No grade MPNST Spindle cell No neoplasm Atypical cells present Noncontributory
Yes
Yes
Inflammatory cyst
No
Yes
Yes
Yes
205
Rare atypical cells
Noncontributory
Amyloid with tissue reaction
Yes
Yes
Yes
249
Necrotic neoplasm
Yes
Malignant neoplasm with neuroendocrine features
Yes
Dedifferentiated liposarcoma High grade sarcoma with focal myogenous diff. Amyloidoma with plasma cell dyscrasia Ewings sarcoma/ PNET
Yes
196
Dedifferentiated liposarcoma High grade sarcoma
Yes
Yes
No
Summary: 10 of 14 open biopsies following FNA had diagnostic utility. 7 of 14 open biopsies following FNA had clinical utility.
482
Diagnostic Cytopathology, Vol. 42, No 6
Yes Yes
Diagnostic Clinical utility? utility?
Myxoid liposarcoma Small blue cell neoplasm Insufficient
67
Yes Yes
OB OB is significant correct error
Yes Yes Yes
Yes
Yes Yes
Diagnostic Cytopathology DOI 10.1002/dc
??? ???
Yes Yes
4 of 14 open biopsies following CNB had diagnostic utility. 3 of 14 open biopsies following CNB had clinical utility. 2 of 14 open biopsies following CNB had diagnostic and clinical disutility.
Yes No Yes Yes Myxoid liposarcoma Cartilage forming neoplasm No e/o neoplasm Malignant poorly diff. neoplasm 160 199 205 274
Yes Yes Yes Yes
Yes Leiomyosarcoma Poorly diff. carcinoma 101 136
Yes No
No Yes
Myxoid and round cell liposarcoma Chondroblastic osteosarcoma Amyloid with tissue reaction Necrosis, rare malignant cells
Yes Yes
Yes
Yes
Leiomyosarcoma gr. 3 MPNST with rhabdomyoblastic differentiation (malignant triton tumor) Myxoid liposarcoma Chondromyxoid fibroma Amyloidoma with plasma cell Dyscrasia Angiosarcoma
Yes Yes Yes
??? ???
Yes Yes
Myxoma Anaplastic large cell lymphoma MPNST High grade pleomorphic sarcoma Synovial sarcoma Dedifferentiated liposarcoma Pleomorphic liposarcoma Myxoid and round cell liposarcoma Yes Yes 1 Yes No Yes Yes Yes Yes
Myxoma Anaplastic large cell lymphoma MPNST Inflammatory pseudotumor Monophasic synovial sarcoma Dedifferentiated liposarcoma High grade sarcoma Myxoid liposarcoma vs. chondrosarcoma Leiomyosarcoma gr. 3 High grade sarcoma Yes
Yes No Yes Yes Yes Yes No No Myxoid neoplasm SBRCT c/w Ewing’s MPNST Spindle cell sarcoma gr.2/3 High grade sarcoma Pleomorphic spindle cell neoplasm Malignant solitary fibrous tumor Small round blue cell tumor
CNBdx Case ID
11 18 25 44 72 76 82 99
CNB Signif Error CNBdx Correct
Table X. Summary of Utility of Open Biopsy following CNB
Open Biopsy Dx
OB dx Is Correct
OB Dx Signif Error
Resection Dx
Diagnostic Utility
Clinical Utility
DIAGNOSTIC ACCURACY AND CLINICAL UTILITY OF BIOPSY
biopsy. The second most frequent pathway was MR followed by open biopsy. MR followed by FNA was the third most frequent diagnostic pathway selected. Figure 3 illustrates the relationship of the various diagnostic procedures with and without the use of MR. The selection of biopsy technique for a musculoskeletal lesion is not a trivial issue and Mankin et al.1 documented the hazards of both core needle biopsy and open biopsy. Both techniques are associated with significant error and complication rates. Many of these caused the optimal treatment plan to be modified. In his multi-institutional study of 329 patients, fully 18% of biopsies were associated with a significant diagnostic error and an additional 10% resulted in a non-representative or technically poor biopsy.1 Mankin also reported that a significant number of patients undergoing core needle or open biopsy had procedural complications, which subsequently altered the treatment plan. Accuracy of a diagnostic technique can be variably defined as diagnostic accuracy or therapeutic accuracy. Diagnostic accuracy specifically documents the number of correct cases as a percentage of total cases. Therapeutic accuracy is clinically more important in that it reports the number of correct cases and insignificant errors as a percentage of total cases studied. Accuracy statistics for core needle biopsy show it to be a highly reliable technique for the identification of musculoskeletal lesions.2–14 Nonetheless, accuracy rates for separating benign from malignant lesions have varied considerably but most studies report accuracies above 90%.4,5,11–14 Based on these statistics, a number of authors have recommended core needle biopsy as an alternative to open biopsy.3,6,7 In many institutions including our own, core needle biopsy is performed on the majority of patients and is frequently the definitive biopsy technique used. More recently, FNA has been used in the diagnostic pathway for the investigation of musculoskeletal lesions. FNA is associated with good diagnostic accuracy15–26 and in some reports, has a diagnostic accuracy above 90%.21,25,26 However, some surgeons remain reticent to use FNA as the definitive diagnostic technique. Most studies directly comparing FNA and core needle biopsy have demonstrated a slight superiority in diagnostic accuracy for core needle biopsy.27–29 Logan et al.30 developed a biopsy algorithm using FNA and core needle biopsy. Logan’s algorithm yields an overall diagnostic accuracy of 96%.30 Our study included 257 patients referred to a musculoskeletal center. The majority of patients had soft tissue lesions (97%) and most were malignant (77%). The study protocol required a resection specimen for inclusion and this skewed the study population toward malignancies. At the time of the study, no protocol existed to determine which biopsy techniques were used and in what order. The
Diagnostic Cytopathology, Vol. 42, No 6
483
Diagnostic Cytopathology DOI 10.1002/dc
LAYFIELD ET AL. Table XI. Factors Affecting the Probability of an Open Biopsy Factor FNA CNB concordance MR CNB concordance MR FNA concordance Whether MR was performed Whether FNA was performed Whether CNB was performed MR Diagnostic category CNB Diagnostic category FNA Diagnostic category
Conclusion
P value
Not significant Not significant Not significant Significant Significant Significant Significant Not significant Not significant
0.25 0.39 0.18 0.012
