All cases of DFSP included in the pathological database of Yonsei University Health System in Seoul, Korea from 1997 to 2012 were reviewed. Informed consent was obtained in accordance with the ethical committee procedures of the institution (IRB No. 4-2011-0659). The following clinical data were recorded: age, gender, tumor location, tumor size (centimeters), tumor type (primary or recurrent), surgical treatment by WLE or MMS, number of micrographic stages, type of surgical repair (primary closure, graft, or flap), local recurrence, distant metastasis, and follow-up period (months). Patients with both primary and recurrent lesions were included. Primary disease was defined as the presence of tumor without previous treatment or recent tumor excision with histologically positive margins. Recurrent disease was defined as that occurring within or juxtaposed to the previous surgical excision site ≥6 months after the initial excision. Histological confirmation of DFSP was done before WLE or MMS in all patients, and all tumors were stained for CD34. Patients were contacted to evaluate the surgical sites for recurrence. Data on recurrence and any subsequent procedures were collected and compiled.

All patients were treated by one of the two surgeons. Patients were treated by WLE or standard MMS using the fresh frozen tissue technique. WLE was performed with a 2-cm margin. For MMS, the visible border of each lesion was demarcated, and a 5-mm margin was drawn around each lesion's clinical edge. In cases of recurrence, scars were removed en bloc with the recurrent tumor before the MMS layer was excised to identify any residual tumor around the scar tissue. Afterwards, the tissue sections were placed in cassettes and submitted with the MMS map for pathological examination. If any evidence of tumor was seen within the specimen, further tissue removal was performed. Additional 5-mm-thick layer of tissue was removed after careful mapping of the positive margin, and this layer was pathologically examined as described above. This process was repeated until all margins were cleared. All sections were embedded in paraffin wax for later confirmation by haematoxylin and eosin (H&E) staining, and CD34 staining was also used in suspicious cases. For all patients, the central bulk of the tumor was sent to the pathology laboratory to have permanent sections taken to confirm the diagnosis of DFSP. All patients underwent primary reconstruction by the two surgeons.

Immunohistochemistry was performed on paraffin-embedded archival tissue with antibodies against CD34 (1:50; BioGenex, San Ramon, CA, USA), p53 (1:50; Dako, Carpinteria, CA, USA), and Ki-67 (1:50; Dako, Carpinteria, CA, USA). Appropriate positive and negative controls were run in parallel. Immunostaining was evaluated separately by two authors. Quantification of the immunohistochemistry marker CD34 was based on the percentage of immunoreactive cells as follows: high (≥50%) or low (<50%). The immunohistochemical quantification of nuclear markers was evaluated as follows: positive (>5% immunoreactive nuclei) or negative (<5%).

Total RNA was extracted from formalin-fixed, paraffin-embedded tissue blocks using an RNeasy Mini Kit (QIAGEN, Hilden, Germany) and reverse-transcribed using a Superscript Preamplification System (Gibco-BRL, Gaithersburg, MD, USA). To detect the expression of the PDGFB gene, first-strand cDNA was amplified with the AGexpdgf3 and AGexpdgf4 primers according to the methods of Greco, et al.10 To detect the presence of COL1A1-PDGFB fusion transcripts, polymerase chain reaction (PCR) was carried out using 17 COL1A1 forward primers and a specific PDGFB reverse primer, as presented in Table 1. Seventeen COL1A1 forward primers were designed for COL1A1 exons 5, 8, 11, 15, 17, 20, 23, 26, 27, 32, 35, 38, 40, 42, 44, 46, and 49, and these primers were considered sufficient to span the various breakpoints within the region encoding the alpha-helical domain of the COL1A1 polypeptide. The PCR products were directly sequenced using an Applied Biosystems 373A automated DNA sequencer to identify the breakpoints.

Associations between the clinical, histopathological, and genetic aspects were assessed using a χ² test and Fisher's exact test. The significance level was set at 5%. All tests were performed using a statistical software package (SPSS, Version 18.0, SPSS Inc., Chicago, IL, USA).

In total, 37 cases of DFSP were reviewed. The clinicopathological features are summarized in Table 2. The mean age at diagnosis was 37.4 years, and the ratio of males to females was 1.1:1. The most common tumor location was the trunk (43.2%), followed by the extremities and the head and neck. Of the 37 cases analysed, 24 cases (64.9%) were de novo neoplasms and 13 (35.1%) were recurred cases. All the recurred cases were previously treated by simple excision or WLE. The median preoperative tumor size was 6.2 cm². All patients were treated primarily by surgery: 34 (91.9%) cases by MMS and 3 (8.1%) by WLE. One patient with recurred DFSP was treated with imatinib mesylate as a neoadjuvant therapy before MMS, because of multiple recurrences and the location of the tumor, which was on the finger web.11 Imatinib 400 mg twice a day for 3 months followed by reduced dose, 400 mg once a day for 2 months, was prescribed. MMS was performed after reduction of tumor size with Imatinib treatment. For those tumors that were treated with MMS, a mean of 1.8 Mohs stages were required. The median postsurgical defect size was 19.6 cm². Postsurgical defects were reconstructed by primary closure in 23 cases (62.2%), covered by flaps in 9 cases (24.3%), grafts in 3 cases (8.1%), and secondary intention healing in 2 cases (5.4%). None of our patients were treated with radiotherapy after surgery. The mean follow-up period was 33 months (range 6-79), and 2 patients (1 MMS and 1 WLE) demonstrated local recurrence during follow-up. None of the patients showed distant metastasis. Of the 34 cases of DFSP treated by MMS, 1 case recurred. The permanent pathological slides and CD34 stain of the MMS specimens of this patient were reviewed afterwards, which showed positive margins.

H&E-stained slides were reviewed. The tumors were categorized histologically as conventional DFSP versus special variants: giant cell fibroblastoma (GCF, combination of spindle cell patterns with myxoid areas, multinucleated giant cells, and distinctive sinusoid-like spaces), myxoid (DFSP with >50% myxoid stromal changes), or the presence of areas with high-grade fibrosarcomatous changes (DFSP-FS) in at least 5% of the lesion. The high-grade fibrosarcomatous areas could be recognized by fascicular, herringbone growth patterns at low power and unusually increased cellularity and cytologic atypia.12 Conventional DFSP (28 cases, 75.7%) was the most common histologic type in our study. In 5 cases (13.5%), storiform pattern was admixed with high-grade cellular areas to form a herringbone appearance which is consistent with DFSP-FS. There were 2 cases (5.4%) of myxoid DFSP and 2 cases (5.4%) of GCF. In 19 cases (51.4%), tumors showed tentacle-like projections into the underlying subcutaneous tissue, resulting in a honeycomb appearance. In 8 cases (21.6%), the tumor showed muscle infiltration. Compared with conventional DFSP, DFSP-FS showed significant subcutaneous (p<0.05) and muscle infiltrations (p<0.001). The histologic subtype of the 2 patients who showed recurrence in our study was DFSP-FS. The recurrence rate of DFSP-FS was significantly higher compared to other types of DFSP (p=0.015).

All of the tumors expressed CD34. In our study, all DFSP-FS demonstrated high CD34 expression. However, there was no significant difference in CD34 expression compared with conventional DFSP. Ki-67 and p53 were expressed in 27.3% and 24.2% of cases, respectively, and were more prevalent in DFSP-FS than in conventional DFSP (Table 3). All 5 cases of DFSP-FS showed positive immunostaining for Ki-67, compared with 14.3% positivity in conventional DFSP (p=0.003). For p53, 80% of the cases in DFSP-FS stained positive, compared with 14.3% in conventional DFSP (p=0.023).

In 16 of 37 cases, reverse transcriptase-PCR analysis was performed. Only 16 cases out of 37 contained human RNA of sufficient quality/quantity for PCR amplification. The COL1A1-PDGFB fusion gene was expressed in 14 (87.5%) cases. Sequencing analysis of the different fusion types revealed a high variety of combinations among several exons of COL1A1 (from exon 5 to exon 49) and exon 2 of PDGFB. The COL1A1 breakpoints occurred at the following exons: 25 (two cases), 29 (two cases), 32 (two cases), 34 (five cases), 43 (one case), and 45 (two cases). No association was found among the different COL1A1-PDGFB fusion transcripts, histological subtypes and clinical features. A schematic representation of our findings is depicted in Fig. 1.