Limb volume measurement for chemotherapy dosing in hyperthermic isolated limb perfusion using serial circumferential limb measurement (A) with computer modeling (B).
Response to hyperthermic isolated limb perfusion (HILP) in 58 cases with evaluable data. CR indicates complete response; NR, no response or progression of disease; and PR, partial response. Responses are described in the “Methods” section.
Distant disease-free survival rates of patients with a positive vs a negative lymphadenectomy finding. Differences were statistically significant by the Kaplan-Maier method and log-rank test.
Overall survival rates of patients with a complete response (CR) to therapy vs a partial response or a nonresponse. Differences were statistically significant by the Kaplan-Maier method and log-rank test. Responses are described in the “Methods” section.
Aloia TA, Grubbs E, Onaitis M, Mosca PJ, Cheng T, Seigler H, Tyler DS. Predictors of Outcome After Hyperthermic Isolated Limb PerfusionRole of Tumor Response. Arch Surg. 2005;140(11):1115-1120. doi:10.1001/archsurg.140.11.1115
Copyright 2005 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2005
Analysis of multiple clinical and pathological factors in patients undergoing therapeutic hyperthermic isolated limb perfusion for extremity melanoma can identify variables with prognostic significance.
Retrospective review of a prospectively collected limb perfusion database with a median follow-up interval of 32.2 months.
Single-institution tertiary care surgical oncology unit.
We report a series of 59 consecutive therapeutic hyperthermic isolated limb perfusion treatments (14 upper extremity and 45 lower extremity) in 54 patients with melanoma from January 1, 1995, through December 31, 2002, using a standard melphalan dosing protocol. At the time of perfusion, 31 cases had fewer than 10 lesions, with none greater than 3 cm in diameter. The remaining 28 cases had 10 or more lesions or at least 1 lesion greater than 3 cm in diameter.
Main Outcome Measures
Response, recurrence, and survival were assessed in relation to multiple demographic, clinical, and technical variables using χ2, log-rank, and Kaplan-Meier survival analyses.
The 3-year survival for the entire cohort was 54%. Thirty-three (56%) of the 59 perfusion treatments resulted in a persistent complete response of at least 6 months’ duration. Statistical analysis showed that patients with no evidence of regional nodal involvement had a significantly lower incidence of distant recurrence (P = .02). Those patients achieving a complete response to therapy had a survival advantage (P = .03).
In patients undergoing therapeutic hyperthermic isolated limb perfusion for in-transit melanoma, the ability to achieve a complete response following treatment, independent of regional nodal status, was the strongest predictor of long-term survival.
The most common indication for hyperthermic isolated limb perfusion (HILP) is the treatment of patients with in-transit extremity melanoma, where the pattern of disease contraindicates a limb-sparing wide local excision. When used in this setting, complete response (CR) rates of 50% to 60% can be expected, and rates as high as 82% have been reported.1,2 Despite HILP response rates that are significantly higher than those obtained with adjuvant systemic chemotherapy,3 no study has demonstrated a survival benefit for therapeutic HILP over alternative therapies in patients with melanoma.
Although a survival benefit of HILP has not been established in randomized trials, there appears to be a subset of patients with in-transit melanoma who have an indolent course. This subset may achieve long-term disease-free intervals following aggressive regional therapy.4 To identify this subset of patients, a number of studies have examined clinical factors that appear to be associated with improved long-term outcome in patients undergoing therapeutic perfusion. These studies commonly report that younger age, shallower depth of primary melanoma, and negative lymph node status are predictors of improved outcome following HILP.5- 7 To date, the influence of initial response to therapy on long-term outcome has not been established.
The goal of this study was to examine our early experience with therapeutic HILP for extremity melanoma to identify treatment-related factors that may be associated with toxic effects, response, recurrence, and survival.
A review of our prospectively collected melanoma surgical database identified 54 patients who underwent 59 therapeutic HILP treatments from January 1, 1995, through December 31, 2002. To treat recurrent disease in the same limb, 3 patients underwent a second perfusion, and an additional patient underwent reperfusion twice. In all cases, the indication for HILP was locally advanced or in-transit melanoma with a distribution and/or disease burden that contraindicated local excision. Preperfusion tumor burden was measured and recorded. Patients with fewer than 10 tumor nodules and no nodule greater than 3 cm in diameter were classified as having low-burden disease, whereas patients with 10 or more tumor nodules or at last 1 nodule greater than 3 cm in diameter were classified as having high-burden disease.
Chemotherapy dosing was based on limb volumetrics, which were calculated by the addition of limb circumference measurements made at 1.5-cm intervals for the length of the limb (Figure 1). The distribution of perfusion sites included 45 lower extremity and 14 upper extremity perfusions. During the course of 42 of the 59 perfusions, a lymph node dissection was performed. Patients were selected for concurrent lymphadenectomy if the cannulation was performed through a previously undissected regional nodal basin or if clinically evident nodal disease was present.
Surgical technique included isolation of the limb circulation, arterial and venous cannulation, and connection with a membrane oxygenation circuit. Before chemotherapy infusion, we confirmed vascular isolation by less than 1% leak of radiotracer to the systemic circulation. Melphalan was subsequently administered at a dose of 10 mg/L of limb volume for lower extremity perfusion and 13 mg/L for upper extremity perfusion. The duration of chemotherapy perfusion was 1 hour, followed by a 15-minute washout with isotonic sodium chloride solution. We measured limb temperatures with intramuscular probes and used warming blankets and perfusate fluid warmers for a tightly controlled limb temperature that ranged only from 38.5°C to 40.0°C.
All patients undergoing HILP for extremity melanoma at our institution were entered into a prospective, institutional review board–approved database. Preoperative data collection included the patient’s age, sex, site of melanoma, tumor burden, histological characteristics, and previous therapies. Perioperative variables recorded included the percentage of systemic leak, the total melphalan dose, and whether or not a node dissection was performed. Recorded postoperative outcomes included toxic effects, response, local recurrence, distant recurrence, and death. The development of early postoperative toxic effects was recorded and graded after each perfusion using the 5-level HILP toxicity grading system developed by Wieberdink and colleagues8 in 1992. In addition, for those patients who underwent lymphadenectomy, results of pathological analysis of the lymph nodes was recorded.
Response to therapy was graded as complete, partial, or a nonresponse. Complete response was defined as a perfusion that resulted in a patient with no evidence of clinical disease at 6 months after perfusion. Partial response was defined as a perfusion that resulted in a 50% or more reduction in gross tumor burden as measured at 6 months after perfusion. Nonresponse was defined as a reduction in gross tumor burden of less than 50% at the 6-month postperfusion time point or as disease progression.
The 3 long-term outcomes measured included limb disease-free survival (DFS), distant DFS, and overall survival. Limb DFS was calculated by the time from perfusion to the first local recurrence. Distant DFS was calculated as the time from perfusion to the first distant recurrence. Overall survival was calculated as the time from perfusion to death. In patients who did not reach these end points, follow-up times were censored and calculated from the date of perfusion to the date of last clinical follow-up.
We used the χ2 test to statistically analyze the relationship between study variables and discrete postoperative outcomes (eg, response). Relationships between study variables and time-dependent end points (eg, overall survival) were determined by the Kaplan-Meier method and compared using the log-rank test. We performed all statistical analyses with the STATISTICA software program (version 6.1; StatSoft, Inc, Tulsa, Okla).
More than half of the perfusions were associated with no or only minor toxic effects. Forty-four percent of perfusions resulted in toxic effects of grade III or higher. Most advanced toxic effects were classified as grade III on the basis of persistent range-of-motion or other sensorimotor deficits. Study variables associated with the development of advanced regional toxic effects included age younger than 60 years (P = .005) and female sex (P = .03).
In our cohort, a total of 41 perfusions were accompanied by regional lymph node dissection. In 22 cases (54%), pathological analysis showed nodal metastases. The 3-year distant DFS for patients with a negative lymphadenectomy finding was 51%, whereas only 26% of patients with positive regional nodes were free of distant metastasis at 3 years (P = .02).
Thirty-three (57%) of 58 patients with evaluable results had a CR following HILP. In 30 of these cases, perfusion resulted in the absence of clinical disease at 3 months after perfusion that was sustained at the 6-month postperfusion time point. In 3 additional cases, perfusion led to the reduction of clinically apparent in-transit disease to a single focus at 3 months after perfusion. All 3 patients underwent excision of these foci, and pathological analysis of the excision specimen indicated no residual melanoma present. Following excision, all 3 remained disease free at 6 months after perfusion and were included in the group with CR. Eighteen perfusions (31%) resulted in partial responses, and 7 (12%) were classified as nonresponses (Figure 2). A number of preoperative and perioperative variables were analyzed for their relationship to response within the limb, and no statistically significant correlations between this set of variables and response to perfusion were seen (Table 1).
During follow-up, 11 (33%) of the 33 cases with a postperfusion CR developed local recurrence. The median time interval between HILP treatment and development of local recurrence was 13.4 months (range, 6.9-44.8 months). In the 22 patients without local recurrence, the median follow-up interval was 24.7 months. A number of preoperative and perioperative variables were analyzed for their relationship to local recurrence, and no statistically significant correlations were identified (Table 2).
During follow-up, 28 cases (47%) developed distant recurrence. The most common sites of distant recurrence (in descending order) were the lung, skin, gut, and brain. The median time from HILP to distant recurrence was 7.5 months (range, 1.5-49.7 months). For patients free of distant metastases, the median follow-up interval was 32.5 months. Univariate analysis of multiple preoperative and perioperative variables showed that pathological evidence of regional nodal involvement was associated with a significantly shortened distant DFS (P = .02; Table 3 and Figure 3). Patients with regional nodal metastasis had a median distant DFS of only 10.8 months, whereas more than 50% of patients without regional nodal involvement were free of disease at the time of data analysis.
Overall, 3-year survival was 54%, with a median follow-up interval of 32.2 months for patients alive at the date of last follow-up. Of the 20 patients who died during the follow-up period, 19 deaths were attributed to malignant melanoma. In statistical analysis of study variables, the ability to obtain a CR was the strongest predictor of survival (P = .03; Table 4 and Figure 4). Patients with a CR following HILP were more than twice as likely to be alive 3 years after perfusion (67% vs 28%).
In this report we present our early experience with therapeutic HILP in the treatment of extremity melanoma. Multiple preoperative and perioperative factors were analyzed for their impact on outcome measures, including toxic effects, response, recurrence, and survival. We found that our level of toxic effects was acceptable and that our CR rate of 57% was consistent with response rates reported by other institutions.9,10 Analysis of our data revealed 2 factors with prognostic significance. First, patients with regional nodal metastasis at the time of HILP were more likely to develop subsequent distant metastases. Second, we found that the ability to obtain a CR following HILP was the strongest predictor of survival.
The finding that a CR to chemotherapy predicts patient outcome is seen in a number of malignancies, including esophageal, gastric, lung, breast, urothelial, and colorectal cancer.11- 17 In these instances, the prognostic significance of the local CR is likely a marker for the chemosensitivity of occult distant metastases. Because HILP does not address occult systemic disease, the most plausible explanation for favorable outcomes in patients with a postperfusion CR is that the perfusion eliminates locoregional disease that otherwise would have served as a future source for systemic metastasis. Alternatively, the finding that a CR is a dominant predictor of prolonged survival may suggest that a subset of in-transit melanoma metastases have a favorable tumor biology. In particular, sensitivity to melphalan-based HILP may correlate with a lower potential for the development of distant metastasis.
The apparent variability in chemosensitivity among various tumors, including melanoma, is likely related to genetic factors. Recent work reported by our group suggests that expression of the tripeptide glutathione, which binds to and inactivates melphalan, and the enzyme glutathione S-transferase, which accelerates the binding between melphalan and glutathione, is important in determining the chemosensitivity of melanoma cells to melphalan.18
Another factor that may influence response rates following HILP is melphalan pharmacokinetics. Plasma melphalan concentration during HILP is determined not only by total melphalan dose but also by the volume of perfusate within the circuit. This finding has been confirmed by studies using dye dilution techniques that demonstrate a correlation between circuit volume, peak melphalan concentration, and toxicity.19 All patients in our study received a dose of melphalan based on a uniform protocol thought to provide a consistent melphalan dose per liter of tissue perfused. Despite a standard method for limb volume measurement and the consistency of our dosing regimen, we have shown that melphalan plasma concentrations in our patients varied widely.20 This study suggested that second-compartment drug concentrations and volume of distribution might be extremely important determinants of regional treatment response and toxic effects and that the contribution of pharmacokinetic variables to response and toxic effects should be reexamined.
In addition, our results highlight the opportunity available to develop new strategies for the treatment of patients with locally advanced and in-transit melanoma of the limb. To improve treatment response and minimize toxic effects, some of these strategies might include resistance modulation for melphalan and changes in current melphalan dosing schemes for certain patient groups, including women and those undergoing reperfusions. In addition, the prognostic significance of a CR to perfusion potentially allows response to be used as a surrogate marker for the efficacy of novel regional treatment approaches to advanced extremity melanoma.
Correspondence: Douglas S. Tyler, MD, Division of Surgical Oncology, Department of Surgery, Duke University Medical Center, DUMC Box 3118, Durham, NC 27710 (firstname.lastname@example.org).
Accepted for Publication: December 21, 2004.