3.1. Patient Demographics
A total of 118 patients with unilateral upper extremity secondary lymphedema were included in the study (
Table 1). The average age was 54 ± 11 years, while the average BMI was 26.1 ± 3.9 kg/m
2. Most patients (
n =116; 98.3%) were women, and the most common cause of lymphedema was breast cancer treatment (
n = 96, 81.3%). The average duration of lymphedema was 41 ± 54 months, and the average time to develop lymphedema following surgery was 23.8 ± 57.1 months. The vast majority of patients had International Society of Lymphology (ISL) stage 1 or 2. Briefly, ISL staging is as follows: stage 0: subclinical lymphedema, 1: mild lymphedema that improves with elevation, 2a: moderate pitting edema, 2b: non-pitting lymphedema, and 3: elephantiasis and irreversible skin changes.
The most common referral source was self-referral (35.6%), followed by breast surgeon (21.2%), other plastic surgeons (19.5%), oncologist (11.9%), and physical/occupational therapist (6.8%). The most common compression regimen frequency was around the clock compression (n = 55, 47%); 12 (10%) patients did not use any compression therapy.
The most common signs and symptoms of lymphedema at the time of evaluation in our series is presented in
Figure 1. Swelling was reported by 115 patients (97.5%): Swelling of the hand by 68 patients (57.6%); forearm and elbow by 99 patients (83.9%); upper arm by 99 patients (83.9%). Fifty-six patients (47.5%) reported swelling of the entire upper extremity. Forty patients had a history of cellulitis of the affected extremity (33.9%); of these, 5 (4.2%) were on prophylactic antibiotics at the time of evaluation.
Other symptoms included heaviness (n = 84, 71%), pitting edema (n= 84, 71%), history of infections (n = 40, 33.9%), pain (n = 29, 25%), and anxiety (n = 28, 24%). Common findings on the clinical exam, other than swelling and pitting edema, included upper extremity sensory impairment (n = 19, 16%), motor impairment (n = 10, 9%), and shoulder joint limited range of motion (n = 45, 38%).
3.2. Limb Volume Measurements Are More Sensitive and Specific than Circumferential Measurements
Because limb swelling was the most common manifestation of lymphedema, a considerable amount of time was spent evaluating the most common techniques being used to quantify this symptom. There is significant heterogeneity in the literature on this topic: Some groups measure limb volume using a perometer, some use manual circumferential measurements and derive limb volume using a truncated cone formula, and others quantify swelling using circumference measurements alone [
28].
The average percent increase in limb volume in our patients based on ISL staging was as follows: Stage 0 = +2.3 ± 2.3%; stage 1 = +5.6 ± 8.5%; stage 2 = +26.2 ± 19.1%; stage 3 = +36.2 ± 16.4% (
Figure 2A). The difference between stage 1 and 2 was statistically significant (
p < 0.001). However, there were only 2 patients in this series with ISL stage 0 or stage 3 lymphedema. Limb volume and ISL stage were positively correlated, but this correlation was very weak (r
2 = 0.235). Stated differently, limb volume increases were only responsible for approximately 23.5% of the variance in lymphedema stage. This data suggests that ISL staging is not granular enough to stratify patients with a diagnosis of lymphedema if limb volume is considered the most important presenting symptom.
Consistent with previous studies, we found that perometer measurements were highly correlated with volume measures derived from circumferential measurements using the truncated cone formula (r
2 = 0.814,
p < 0.001;
Figure 2B) [
27]. However, manual measurements tended to underestimate the total volume as compared to the perometer (
Figure 2C). The average limb volume assessed by perometer measurement was 2,466 ± 451 mL while the average for manual measurement was 2,216 ± 433 mL. Although this difference was not statistically significant, this finding suggests that measurements should be performed consistently using the same method over time and that volume measurements obtained with the perometer are not interchangeable with those derived from manual circumferential measurements.
Perometer measurements of limb volume difference were also correlated with upper (32 cm above the wrist) circumferential measurements (r
2 = 0.705,
p < 0.001;
Figure 2D). This finding is not surprising given the high degree of correlation between manual measurements and the perometer. However, it is likely that single measurements are more prone to error. To determine the sensitivity, specificity, and positive predictive value of single (12 cm proximal to the wrist) or dual-site (12 cm and 32 cm proximal to the wrist) circumferential measurements, the commonly used >2 cm difference was used as diagnostic of lymphedema. The >2 cm circumference difference was then compared to the commonly accepted diagnostic threshold for limb volume difference by >10% assessed by perometer. This analysis revealed that circumferential measurements had a relatively low sensitivity (82.8%) and specificity (85.3%) as compared with limb volume measurements. In fact, circumferential measurements had only a 74.4% positive predictive value, suggesting that using circumference measurements alone tends to under-diagnose and under-estimate the degree of lymphedema (
Table 2)
3.4. Even Minor Changes in Limb Volume can Have Significant Effects on PRO Measures
Forty-nine patients completed the LLIS, and of these, 42 patients also completed the ULL-27 questionnaire. In both tests, a score of 0 is equivalent to no impairment, while a score of 100 is consistent with complete impairment. The average percent impairment reported the LLIS was 40.7 ± 20.7% for the physical, 35.5 ± 22.1% for the psychological, and 34.8 ± 19.5% for the functional scales. The overall impairment score for the LLIS was 37.2 ± 18.8%. Percent impairment scores for the ULL-27 were 39 ± 24% for the physical domain, 40 ± 23% for the psychological domain, and 26 ± 22% for the social domain (
Table 3). The overall impairment score for the ULL-27 was 36.8 ± 20.3%.
The correlation between the individual domains (Physical: r
2 = 0.657,
p < 0.001; Social/functional: r
2= 0.534,
p < 0.001; Psychological; r
2 = 0.478,
p < 0.001;
Figure 4A–C) was also statistically significant, though these correlations were weaker than the overall impairment. There was a significant correlation between overall impairment as measured by the LLIS and the ULL-27 (r
2 = 0.76,
p < 0.001;
Figure 4D)
. Taken together, these results suggest that patients with upper extremity lymphedema have significant impairment in all domains tested by both tests.
We next analyzed the correlation between LLIS and ULL27 scores and ISL stage (
Figure 5A–F). Patients with ISL stage 2 lymphedema had significantly higher disability scores in all domains as compared with individuals with stage I disease (physical: 45 ± 21% vs. 30 ± 16%,
p = 0.01; psychological: 39 ± 23% vs. 26 ± 16%,
p = 0.03); functional: 39 ± 20% vs. 24 ± 13%,
p = 0.01). We noted a similar pattern in the ULL27 instrument (physical: 46 ± 25% vs. 25 ± 12%,
p = 0.006; psychological: 46 ± 24% vs. 26 ± 14%,
p = 0.008); social: 31 ± 25% vs. 14 ± 11%,
p = 0.02).
We next sought to determine if PRO measures were correlated with changes in limb volume excess (
Figure 6). LLIS physical and function scores were significantly but weakly correlated with limb volume excess (r
2 = 0.24,
p = 0.048,
Figure 6A and r
2 = 0.3,
p = 0.018,
Figure 6C respectively). Stated differently, limb volume changes in our patients were only responsible for approximately 20–30% of the variance in physical and functional disability. In contrast, the correlation between limb volume difference (%) and psychological impairment as measured by the LLIS was not statistically significant (
Figure 6B). In fact, some of our patients displayed significant psychological (as well as physical and functional) impairment, even with modest increases in limb volume. The ULL27 domains had an overall similar slope and pattern of distribution, however, the correlation between limb volume excess and reported measures in physical, psychological, and social domains were not statistically significant (
Figure 6D–F).
Taken together, our findings with PROM in patients with secondary lymphedema of the upper extremity suggest that the LLIS may be more sensitive than the ULL27 for measuring the degree of physical and functional disability resulting from lymphedema limb volume excess. Our findings also show that the correlation between physical/functional impairment and increases in limb volume excess is weak suggesting that even minor increases in limb volume can have a significant impact on the quality of life measures of some patients with upper extremity lymphedema. This hypothesis is supported by the fact that we found no correlation between psychological impairment and increased limb volume difference. These findings suggest that analysis of PRO is an important aspect of preoperative and postoperative measures in patients who are slated to undergo lymphatic surgery.
3.5. Lymphoscintigraphy Has a Low Specificity and Positive Predictive Value for Lymphedema Diagnosis but May Be Useful for Surgical Planning
Seventy-nine patients (67%) underwent lymphoscintigraphy as part of their initial evaluation in our center (
Table 4). Axillary lymph node uptake was absent in the affected limb in 53 (67%) patients, and delayed uptake was documented in 26 (33%) patients. Dermal backflow was present in 24 (30%) patients.
The validity measures of any pathological findings in lymphoscintigraphy (i.e., lack of axillary uptake or the presence of dermal backflow) are presented in
Table 2. The sensitivity of lymphoscintigraphy to diagnose lymphedema in patients with a minimum limb volume excess of 10% was 88% and the positive predictive value was 72.1%. However, the specificity and negative predictive value of lymphoscintigraphy were low (41.4% and 66.7%, respectively). These results suggest that lymphoscintigraphy alone is insufficient in some cases to diagnose lymphedema. However, this test may have utility in identifying patients who have axillary lymph node uptake thus guiding the degree of scar excision and axillary dissection that may be performed safely in the course of a vascularized lymph node transplant. Thus, in patients with axillary lymph node uptake is demonstrated by lymphoscintigraphy, a more prudent axillary scar excision may be necessary to prevent injury to the remaining functional lymphatic vascular network.
3.6. ICG Lymphography Is Highly Sensitive for Detecting Lymphedema
Ninety- seven patients (82%) underwent outpatient ICG lymphography as part of their lymphedema initial evaluation (
Table 4). Every patient in this study had a presence of a pathologic ICG pattern in the limb with lymphedema. Stage 1 (Splash pattern) was observed in 20 patients (20.6%), stage 2 (Stardust pattern) in 62 patients (63.9%) and stage 3 (diffuse) in 15 patients (15.5%). In contrast, all normal limbs in this study had a normal linear ICG pattern.
Both L-Dex score and limb volume did not significantly correlate with ICG stage. Patients with the early stage splash patterns on ICG tended to have lower L-Dex scores and lower limb volumes, but aside from this trend, there was no tight correlation (
Figure 7A,B). This supported the authors’ observation that limb volume and related extracellular fluid content do not necessarily reflect the physiology of the abnormal lymphatic system. Interestingly, there were patients with fairly advanced abnormal ICG patterns that had a minimal limb volume difference of less than 10% and an L-Dex score of less than 10. ICG may uncover deeper pathology than appreciated if only assessing volume and bioimpedance, but this is a subject requiring further study.
There was no correlation between ICG staging and changes in patient reported outcomes in either the LLIS or the ULL27 (
Figure 8A–F). The average impairment score for all three domains was not significantly different in patients with splash, stardust, or diffuse patterns suggesting that ICG staging is not predictive of the degree of disability resulting from lymphedema. This may be due to the limitations of the current ICG staging system, which is entirely subjective and not quantitative. The current system only provides a one-time snapshot of the limb. It does not quantify clearance or lymphatic transport which would be a more meaningful measure of lymphatic function. ICG is an effective means for directly visualizing the superficial lymphatic collectors and is an area requiring further investigation.
3.7. MRA Imaging Is Useful for Identification of Abnormalities in Venous Outflow and Diagnosis of Lymphedema
Table 4 summarizes the findings of MRA performed during the initial evaluation. Seventy-eight patients underwent upper extremity MRA (66.1%) and of these, 12 (15.4%) had evidence of narrowing or stenosis in the axillary vein. One patient (0.8%) had a partial thrombus in her axillary vein. Identifying venous pathology is important because it may contribute to limb swelling and may also compromise the results of lymph node transplant or lymphovenous bypass if not corrected.
Evidence of fluid accumulation was found in 64 patients (82%), and fat hypertrophy was found in 60 patients (77%). One patient in this series was found to have occult bony metastasis on MRA. Of note, an additional patient was found to have incidental bony metastasis on CT angiogram but was not included in these results because she did not have an MRA due to a tissue expander in place. Evidence of fluid accumulation or fat hypertrophy on MRA was highly sensitive for the diagnosis of lymphedema as defined by a limb volume excess of ≥10% (94.2% and 96.2% sensitive, respectively). In other words, nearly every patient who had an increase of 10% or more in their limb volume demonstrated fluid accumulation or fat hypertrophy on MRA. However, these findings had a lower specificity for lymphedema (44% and 64%, respectively). This finding suggests that the pathological changes in lymphedema are progressive and that arbitrary cut-offs in changes in limb volume are not reflective of this process. MRA evidence of fluid accumulation also had relatively high negative and positive predictive values (78.6% and 77.8%, respectively). The negative and positive predictive values were even higher in patients with evidence of fat hypertrophy on MRA (84.8% and 88.9%, respectively) suggesting that MRA is also useful for confirming the diagnosis of lymphedema.