Although BIA derived phase angle has been used as a complementary nutritional assessment tool in cancer, we need to be able to choose a specific cut-off level that can help the treating oncologists and clinical nutritionists classify cancer patients as either well-nourished or malnourished. The choice of the cutoff is mandated by the need to closely match the sensitivity and specificity of the traditional nutritional tests. One way of achieving this goal is to evaluate phase angle against a test that has been extensively validated in similar treatment settings. This study was undertaken to investigate the association between BIA derived phase angle as an indicator of nutritional status, and SGA in advanced colorectal cancer.
The sensitivities and specificities considered together for potential optimal cut-off levels of phase angle were found to be modest at best and the test was found to be either too sensitive or too specific. We found that using different cut-off levels for males versus females, and for newly diagnosed patients versus those with progressive disease might be more appropriate as opposed to using single cut-off level for all patients. Interestingly, a phase angle cut-off of 5.9 demonstrated high diagnostic accuracy in males who had failed primary treatment for advanced colorectal cancer. Our findings are consistent with those reported by another group of researchers who evaluated phase angle against SGA in 279 patients undergoing elective gastrointestinal surgery [1
]. The study found a fair overall agreement between SGA and BIA estimates and couldn't obtain an optimal phase angle cut-off with high sensitivity and specificity. The study also suggested different potential cut-offs for men (6.3) and women (5.9) indicating a better balance of sensitivity and specificity. A study conducted in patients with advanced lung cancer stratified the patient cohort by the mean phase angle score of 4.5. Interestingly, patients with phase angle scores less than or equal to 4.5 had a significantly shorter survival than those with phase angle scores greater than 4.5 [7
]. Another study conducted in HIV-infected patients stratified patients into 4 quartiles, with 5.3, 5.9 and 6.5 as the cut-off points. The study found phase angle to be an independent prognostic marker of clinical progression and survival [17
]. In another prospective study of liver cirrhosis patients, phase angle equal to or less than 5.4 was associated with shorter survival as compared to phase angle greater than 5.4 [10
In our study, no optimal phase angle cut-off level with simultaneously high levels of sensitivity and specificity could be identified. There are several potential explanations of these findings. 1. BIA derived phase angle is not a valid indicator of nutritional status in advanced cancer, 2. Phase angle and SGA capture different aspects of nutritional status and might complement each other in overall nutritional evaluation, 3. Phase angle is a valid marker of nutritional status and the relatively modest correlations observed in the present study might be escalated using a larger sample size. In the present study a phase angle cut-off level of 5.2 had low sensitivity but high specificity whereas a cut-off level of 6.0 had high sensitivity but low specificity. It is likely that an optimal phase angle cut-off level is located somewhere between these two values. We believe that the goal of achieving an optimal phase angle cut-off with high levels of sensitivity and specificity should be further explored in similar patient populations with larger sample sizes.
ROC analysis at best provides guidelines for which cut-offs should be considered. We believe that the choice of an optimal cut-off level for any diagnostic test is context dependent. In our study, we evaluated the optimal cut-off levels of phase angle as a nutritional assessment tool in advanced colorectal cancer. Since malnutrition is a major cause of morbidity and mortality in these patients, the treating oncologists and clinical nutritionists might find it more worthwhile to be able to correctly identify a high proportion of malnourished patients (a high sensitivity) even though it comes at the expense of reduced specificity (a high rate of false positives). In such situations, selecting a high cut-off level of 6 makes more sense as opposed to selecting a low cut-off level of 5.2.
What exactly is phase angle? Some earlier studies have tried to address these questions, albeit in a limited capacity. For instance, Schwenk et al. hypothesized that phase angle could possibly be interpreted as a global marker of malnutrition in HIV infected patients [17
]. In another study conducted on HIV-infected patients, it was argued that phase angle reflects the integrity of vital cell membranes [15
]. In patients with liver cirrhosis, phase angle was speculated to be a marker of clinically relevant malnutrition characterized by both increased extracellular mass and decreased body cellular mass [10
]. In advanced lung cancer, phase angle was speculated to be an indicator of altered tissue electrical properties [7
Limitations of this study relate to the BIA technique, retrospective study design and small sample size. This study, because of its retrospective nature, relies on data not primarily meant for research. The subgroup ROC analyses were based on small sample sizes without accounting for the number of multiple comparisons made in this study. The non-normal distribution of phase angle in our study could be an effect of the small sample size. Despite these limitations, our study provides valuable insights on what might be an appropriate phase angle cut-of level in patients with advanced colorectal cancer. Clearly, there is a need to validate the diagnostic accuracy of phase angle using larger sample sizes in advanced cancer populations.
It has been suggested that the variability of direct bioimpedance measures (resistance, reactance, and phase angle) depends on age, gender, and body mass characteristics of the study population which could possibly limit the extrapolation of the model [24
]. Some other reported limitations of using BIA for assessment of body composition are hydration status and/or major disturbances of water distribution, body position during procedure, ambient air and skin temperatures, recent physical activity, conductance of the examining table, abstinence from alcohol and caffeinated beverages, food consumption and voiding 30 minutes prior to measurement [26
]. Since the original intent of the BIA in this study was to gather estimates of body composition as part of a baseline nutritional assessment in a clinical setting, not all of these factors could realistically be controlled. Patients were free of visible edema or ascites so there was control for obvious overhydration. However, it is to be noted that hydration status in cancer patients are often due to treatment, therefore, just observing obvious signs of hydration may have overlooked changes that could further have influenced measurement validity. Body position was controlled for because all patients were in the supine position in a bed or on an exam table. Air temperature was within a controlled range in our hospital setting. Physical activity was limited in these patients due to the advanced nature of their disease. Food intake was not controlled for in this clinical setting, which may have contributed to a small amount of variability. No assessment of inter-rater reliability of the users of BIA and SGA was made in this study. This bias, however, was minimized by restricting the use of BIA and SGA to well-trained dietitians with an expertise in the use of these clinical techniques. Moreover, BIA was conducted in all patients using the same analyzer. Finally, we believe that future studies should use a more objective method of nutritional assessment in order to derive the definitive cut-offs of phase angle.