Defining lower limb lymphedema after inguinal or ilio-inguinal dissection in patients with melanoma using classification and regression tree analysis.

OBJECTIVE: This study aims to objectively define the criteria for assessing the presence of lymphedema and to report the prevalence of lymphedema after inguinal and ilio-inguinal (inguinal and pelvic) lymph node dissection for metastatic melanoma. SUMMARY BACKGROUND DATA: Lymphedema of the lower limb is a common problem after inguinal and ilio-inguinal dissection for melanoma. The problem is variably perceived by both patients and clinicians. Adding to the confusion is a lack of a clear definition or criteria that allow a diagnosis of lymphedema to be made using the various subjective and objective diagnostic techniques available.
METHODS: Lymphedema was assessed in 66 patients who had undergone inguinal or ilio-inguinal dissection. Nine patients received postoperative radiotherapy. Assessment was performed by limb circumference measurements at standardized intervals, volume displacement measurements, and volumetric assessment calculated using an infrared optoelectronic perometer technique. Comparisons were made with the contralateral untreated limb. Patient assessment of the severity of lymphedema was compared with objective measures of volume discrepancy. Classification and regression tree analysis was used to determine a threshold fractional leg volume or circumference increase above which patients could self-detect volume changes that they reliably considered to indicate lymphedema.
RESULTS: Based on classification and regression tree analysis, both the whole limb perometer volume percentage change > or = 15% and the sum of circumferences (of 6 defined sites along the limb) percentage change > or = 7% performed well overall in predicting moderate or severe perceived swelling (defined as "lymphedema"). Both definitions predicted lymphedema in approximately the same fraction of patients with misclassification rates of 16% and 15%, sensitivity 56% and 50%, specificity 95% and 100%, respectively. Using > or = 15% of whole perometer volume percentage change, 12% of patients with inguinal dissection had lymphedema compared with 23% of patients with ilio-inguinal dissection. Combining both groups, 18% of patients had lymphedema, positive and negative predictive values 82% and 84%. Using the definition > or = 7% of the sum of circumferences percent change, 7% of patients with inguinal dissection had lymphedema compared with 19% of patients with ilio-inguinal dissection (overall 14% had lymphedema, positive and negative predictive values 100% and 82%, respectively). Of the variables assessed, only radiotherapy was significantly associated with predicted lymphedema (OR 12.6; 95% CI 1.7 to > 100; P = 0.001 using whole perometer change > or = 15%; and OR 13.0; 95%CI 1.4 to > 100; P = 0.021 using sum circumference change > or = 7%).
CONCLUSIONS: A whole limb perometer volume percentage change of > or = 15% and increase in the sum of circumferences of the defined points along the limb > or = 7% provide robust definitions of lower limb lymphedema.