br This is first study to use the
This is first study to use the BIVA RXc z-score method to compare body composition in cancer populations, using data from
previously published bioimpedance data. BIVA offers advantages over traditional methods of body composition assessment, due to its non-invasive nature and simplicity. BIVA has methodological advantage over traditional BIA calculations due to its independence of regression equations (which lack accuracy in cancer ). Furthermore, BIVA can facilitate longitudinal assessments to eval-uate body composition changes over time. These properties are useful to evaluate nutrition and hydration in people affected by cancer, who are unable to tolerate more invasive methods of assessment. This research demonstrates the potential to use pub-lished BIVA data for larger analysis.
Comparison with previous work
The only previous study to use BIVA RXc z-scores in cancer was Piccoli et al., 2002 . Piccoli plotted data from the vector point for males with stage IV lung cancer (Toso et al., 2000 ) within the cachexia quadrant (75% tolerance ellipse). Our data builds on Piccoli's study and describes how, in addition to Toso's stage IV lung cancer data, three other populations were also classified as cachectic. This included a lung cancer sample with local and disseminated disease , and two head and neck cancer cohorts
Fig. 4. RXc z-score graph analysis of bioelectrical impedance vector analysis (BIVA) data from studies of patients with cancer. Data drawn from the literature and plotted on the RXc-score graph after transformation of impedance measurements from several disease groups into bivariate Z scores (with respect to the Piccoli 1995 reference 129453-61-8 ). Further details of the equations used for the analysis are available in the appendix.
[21,32,33]). The vectors for the advanced cancer population described by Nwosu et al.  (although plotted within the normal 50% ellipse) were in a similar position to the lung cancer studies by Toso et al., [18,20]. This suggests similarity between these groups (i.e. low muscle mass, with risk of cachexia), even though body composition was classified as normal. Therefore, interpretation of BIVA Rxc z-score data requires consideration of clinical factors in addition to BIVA.
Previous work illustrates how patients with cancer are prone to develop cachexia as their condition progresses [1,36]. However, data about the stage of cancer was only available for two pop-ulations. It is possible that stratification of data by cancer stage may have demonstrated that individuals with more severe cancer were more likely to be cachexic. Furthermore, assessments at different
Bioimpedance RXc z-score data for the included studies. points in the disease trajectory may demonstrate changing body composition over time.
Our data demonstrates that body composition appeared to be related to cancer type, disease severity and gender . For example, females with breast  and gynaecological cancers  had increased cell mass compared to other populations (demonstrated by more superior vector placement). Two factors may explain this difference. Firstly, individuals with breast and gynaecological cancer were comparatively younger than other groups (the mean age for the breast and gynaecological cancer groups were 53 and 60 years respectively, whereas most other populations were aged >60 years). Secondly, these patients were recruited at diagnosis, whereas par-ticipants in other studies were recruited later in their illness.
A limitation of this study is that nutritional screening tools were not used in all studies, which makes nutritional based comparisons difficult. The Subjective Global Assessment (SGA - a simple bedside method of assessing the risk of malnutrition ) was used in the majority studies. However, only one study (Car-doso et al. ) reported body mass index (BMI) data according to the requirements of the European Society for Clinical Nutrition and Metabolism (ESPEN) malnutrition criteria . Therefore, our ability to evaluate how BIVA RXc z-scores relates to nutritional states is limited.
A small number of studies were evaluated in this analysis and the majority of participants included in the studies were from white, European or North American populations, which limits our ability to extrapolate the findings. The under-representation of non-white groups in these studies may be due to various factors, such as language and cultural barriers . Further, as this analysis only included English language studies, it is possible that studies using BIVA in different cultural contexts were excluded.