• sarcoma;
  • response assessment;
  • magnetic resonance spectroscopy (MRS);
  • limb perfusion;
  • recombinant human tumor necrosis factor-α (rhTNFα)



In patients with unresectable soft tissue sarcoma of the extremities, isolated limb perfusion (ILP) has been reported to result in significant tumor regression enabling limb-sparing resection in the majority of patients. However, clinical tumor response as evaluated by imaging and histopathology (extent of tumor necrosis) often differ significantly. The current study was initiated to evaluate prospectively the role of 31phosphorus-magnetic resonance spectroscopy (31P-MRS) in the noninvasive assessment of histologic response in patients treated with ILP.


Thirty-two patients with locally advanced and unresectable soft tissue tumors (sarcoma in 28 patients and bulky melanoma in 4 patients) were treated by ILP with recombinant human tumor necrosis factor-α and melphalan or with cytostatics. 31P-MRS was performed prior to treatment and at regular intervals after ILP until definite tumor resection. Clinical response parameters according to the World Health Organization as well as the histopathologic necrosis rate of the resection specimen were correlated with changes in the energy-rich phosphorous metabolites phosphocreatine (PCR); α-, β-, γ-adenosine triphosphate (ATP); phosphomonoesters (PME); and inorganic phosphate (Pi).


Clinically, 15 of 32 patients (response rate [RR] of 47%) demonstrated a partial response (PR). The ratios of PME/PCR and PME/β-ATP decreased significantly after ILP in comparison with preoperative values (P < 0.001). The changes in the PME/β-ATP ratio were significantly different between clinical responders and nonresponders (P < 0.02) in contrast with the PME/PCR ratios (P < 0.09). Histologic necrosis of > 90% (pathologic (p) PR) was present in 17 resection specimens, 7 of which demonstrated no clinical response. Seven tumors demonstrated a pathologic complete response (pCR). When combining PR, pPR, and pCR (RR of 68%), 31P-MRS was able to predict response with a specificity of 94% and a sensitivity of 68% (P < 0.006, by the chi-square test).


The considerable difference between clinical and pathologic RR after ILP underlines the shortcomings of established response criteria. Utilizing changes in PME/β-ATP ratios, 31P-MRS is a highly specific tool with which to predict histologic response in this setting. This finding may be of major value in those patients in whom the decision to perform a major resection or amputation must be made for local tumor control. Cancer 2002;94:1557–64. © 2002 American Cancer Society.

DOI 10.1002/cncr.10361