Body composition: DXA and bioimpedance in heart failure

Comparison of two bioelectrical impedance devices and dual-energy X-ray absorptiometry to evaluate body composition in heart failure

Alves et al. JHND Early View.



The utilisation of bioelectrical impedance analysis (BIA) in heart failure can be affected by many factors and its applicability remains controversial. The present study aimed to verify the adequacy of single-frequency BIA (SF-BIA) and multifrequency BIA (MF-BIA) compared to dual-energy x-ray absorptiometry (DEXA) for evaluating body composition in outpatients with heart failure.


In this cross-sectional study, 55 patients with stable heart failure and left ventricle ejection fraction ≤45% were evaluated for fat mass percentage, fat mass and fat-free mass by DEXA and compared with the results obtained by SF-BIA (single frequency of 50 kHz) and MF-BIA (frequencies of 20 and 100 kHz).


MF-BIA and DEXA gave similar mean values for fat mass percentage, fat mass and fat-free mass, whereas values from SF-BIA were significantly different from DEXA. Both SF-BIA and MF-BIA measures of body composition correlated strongly with DEXA (r > 0.8;P < 0.001), except for fat mass assessed by SF-BIA, which showed a moderate correlation (r = 0.760; P < 0.001). MF-BIA also showed a better agreement with DEXA by Bland–Altman analysis in all measurements. However, both types of equipment showed wide limits of agreement and a significant relationship between variance and bias (Pitmans’s test P > 0.05), except MF-BIA for fat-free mass.


Compared with DEXA, MF-BIA showed better accuracy than SF-BIA, although both types of equipment showed wide limits of agreement. The BIA technique should be used with caution, and regression equations might be useful for correcting the observed variations, mainly in extreme values of body composition.

New articles on JHND Early View


Glycaemic responses of staple South Asian foods alone and combined with curried chicken as a mixed meal

R. George, A. L. Garcia and C. A. Edwards. Article first published online: 24 MAR 2014 | DOI: 10.1111/jhn.12232

Effect of weight maintenance or gain in a 10 years period over telomere length, sirtuin 1 and 6 expression and carotid intima media thickness

D. Bunout, G. Barrera, M. P. de la Maza, L. Leiva and S. Hirsch Article first published online: 24 MAR 2014 | DOI: 10.1111/jhn.12231

Evidenced-based, practical food portion sizes for preschool children and how they fit into a well balanced, nutritionally adequate diet

J. A. More and P. M. Emmett Article first published online: 24 MAR 2014 | DOI: 10.1111/jhn.12228

Green tea extract and breast cancer

Effects of a green tea extract, Polyphenon E, on systemic biomarkers of growth factor signalling in women with hormone receptor-negative breast cancer

Crew et al., JHND Early View


Observational and experimental data support a potential breast cancer chemopreventive effect of green tea.


We conducted an ancillary study using archived blood/urine from a phase IB randomised, placebo-controlled dose escalation trial of an oral green tea extract, Polyphenon E (Poly E), in breast cancer patients. Using an adaptive trial design, women with stage I–III breast cancer who completed adjuvant treatment were randomised to Poly E 400 mg (= 16), 600 mg (= 11) and 800 mg (= 3) twice daily or matching placebo (= 10) for 6 months. Blood and urine collection occurred at baseline, and at 2, 4 and 6 months. Biological endpoints included growth factor [serum hepatocyte growth factor (HGF), vascular endothelial growth factor (VEGF)], lipid (serum cholesterol, triglycerides), oxidative damage and inflammatory biomarkers.


From July 2007-August 2009, 40 women were enrolled and 34 (26 Poly E, eight placebo) were evaluable for biomarker endpoints. At 2 months, the Poly E group (all dose levels combined) compared to placebo had a significant decrease in mean serum HGF levels (−12.7% versus +6.3%, P = 0.04). This trend persisted at 4 and 6 months but was no longer statistically significant. For the Poly E group, serum VEGF decreased by 11.5% at 2 months (P = 0.02) and 13.9% at 4 months (P = 0.05) but did not differ compared to placebo. At 2 months, there was a trend toward a decrease in serum cholesterol with Poly E (P = 0.08). No significant differences were observed for other biomarkers.


Our findings suggest potential mechanistic actions of tea polyphenols in growth factor signalling, angiogenesis and lipid metabolism.

How to write #6- the results

So, on to the results section of the paper. I have been putting off writing this part of my guide as it is probably the most difficult part of the process to describe. However, when it comes to writing up a research study this is the section that I recommend you write first. The results are the raison d’etre for the paper- they are the component that the reader most wants to see and this is the heart of the writing.


The function of the results

For most journals (though not all- some will combine the Results and Discussion sections) require the results to be an objective presentation of the findings of your research without any interpretation of the outcomes (this is reserved for the discussion). The results should comprise tables and figures that enable summaries of your findings to be presented for the reader to consider and interpret for themselves, and an accompanying text commentary which describes the content of those tables and figures.

Writing a paper is a different process to writing a dissertation or thesis. Generally when I advise on the latter I will tell students that the text of the results should be a full account of the material in the tables and figures which can be understood without looking at those tables and figures, and that the tables and figures should be equally understandable without the text. For a paper though we need to think about word limits and the need for brevity and clarity. The text of your results is there to signpost to the tables and figures and to deliver key highlight messages and help your reader navigate to what you want them to look at. Lengthy description of null findings, or material which is really just there as background context are not necessary. The major function of the results text is to provide additional information that clarifies aspects of the data. Make sure that you refer to each Table and/or Figure individually and in sequence and explain to the reader the most important results that each is showing.


How to go about it

Usually the decision to write a paper has followed the completion of some statistical or other data analysis (in this series I am not really addressing papers which contain qualitative data as this is not my research milieu). This is the point where you will have found something in your study that you consider to be interesting or which has either led you to reject or support your initial hypotheses. With this being the case the perfect place to start putting your results section (and in fact your paper) together is to plan and draw your Tables and Figures.  This requires some decision making in terms of what mode of data presentation you are going to choose.

The choice of Tables of Figures generally goes far beyond being a simple question of taste. Each format has strengths and weaknesses that makes them well suited to particular kinds of data.

  • Tables
    • Are best for presenting large volumes of data. If you have measured say 4 or more variables that are closely related, it makes far more sense to put them all in a table together rather than having 4 + figures.
    • Are ideal for presenting data that is very general and scene-setting. Readers often won’t pore over the detail of your tables, but will need to refer to them for specific pieces of information.
    • Figures
      • Are the best way of showing simple data.
      • Should be reserved for showing your highlights- figures are visually striking and so are the most effective way of presenting the really important pieces of data.

With Tables and Figures planned, you now need to arrange them into a logical sequence which enables you to use them to evaluate your hypothesis. This order will usually be obvious to you but as a basic rule of thumb you would put the paper ‘fodder’ in first. By fodder I mean the very general background material which isn’t that interesting but which has to be there. This might include a table which gives the basic demographic information about your cohort, or data which describes the basic characteristics of your animals or cells in culture. Moving on from there the data will become more specifically related to the hypotheses. Each additional item would add a further layer of complexity to the data set.

For example:

Table 1. Basic description of the cohort

Table 2. Comparison of intervention group and control group, baseline measures

Table 3 Comparison of intervention and control group, follow-up measures.

Table 4. Odds ratios for outcomes.



Table 1. Body and organ weights of animalas

Table 2. Blood glucose, cholesterol and triglyceride concentrations.

Figure 1. Expression of genes in liver.

Figure 2. Histological evidence of hepatic steatosis.

Table 3. Methylation of gene promoters for gene X.


It is unusual for authors to struggle to decide on the order in which Tables and Figures are presented, more usually the issues lie with deciding which mode of presentation to use, as described above, or with presenting too much (as discussed below).

Once the Tables and Figures are drawn, you will need to write legends for each of them. Here there is a rule of thumb which applies to papers and theses alike. Your legends should enable your reader to understand exactly what you are presenting in your Figure/Table without having to look at any other part of the paper. The legend should at the very least contain a title for the Table/Figure and basic statistical information.

For example:

Table 2. Blood glucose, cholesterol and triglyceride concentrations. Serum metabolites were measured in all rats after 10 days of following the protocol. Data are shown as mean±SEM for n=6 observations. * indicates statistical difference between control and test group P<0.05.

If your statistical analysis is particularly complicated (I myself am never averse to a three way ANOVA) then the legend can be packed with information that conveys the outcome of that analysis. You don’t want to be discussing univariate effects and interactions between factors within the text, as this is generally boring (though very important) and breaks up the flow of the paper.

Once the Tables, Figures and their legends are in place you can write the text of the results section. That is relatively easy as you just describe what you see, crafting the text to follow the sequence of data, highlighting the evidence that addresses the hypotheses and research questions that you aimed to test. There is no great art to this, but there are many things that you can do badly. The section below gives an overview of the mistakes that are typically made.


Mistakes to avoid

Including interpretive comments

It is easy to slip up and include some material that should be in the discussion within your results. For example:

The weight loss observed in the intervention group was significantly greater than in the control group. A loss of 4.8 kg was similar to that reported by Smith and Jones (2013) and was consistent with our initial hypothesis.

Lose the section in red and change this to:

The weight loss observed in the intervention group (4.8kg) was significantly greater than in the control group (0.8kg).

Sometimes it can be difficult to adopt the rather terse and focused reporting style that the results needs, but just focused on brevity and the need to discuss highlights.


Being blinded by your stats-goggles

Many of us become unnaturally obsessed with the outcome of statistical tests when describing our data and I call this the stats-goggles effect. What I mean by this is that the focus switches entirely to statistical significance and moves completely away from thinking about what significance might mean biologically. For example, we might write this sentence,

Weight in the intervention group was significantly different to the control group (P<0.01) and there was an interaction between the effect of intervention and sex (P<0.05, Table 2).

This is meaningless gobbledygook by any definition and does not tell your reader what you observed. Remember that the task in the results section is to say what you observed. The statistical analysis is just a tool to decide which bits of your observations have come about for reasons other than random chance. In describing the data you need to give more information. Think about embellishing the statements you make with the following points:

Directionality: You have a significant difference between groups, but what is the nature of that difference. Is the value measured in the test group bigger or smaller than in the control group? Is there a way of expressing this that makes the sentence more interesting than just a report of statistical fact.

Magnitude: How great a difference have you detected? It may be useful to your reader to report the difference between mean values between two groups, or maybe you could report it as a percentage or fold-difference if that is appropriate. Mix up the way in which you report the magnitude of differences as you move through the results section so that it doesn’t get repetitive.

Relevance: Sometimes we might see an effect that is statistically significant but when we look at the actual mean values that were determined, the difference is not of genuine importance from a biological or clinical perspective. For example if a treatment has lowered total cholesterol concentrations by around 0.1 mmol/L (2% of the range of normal values), is that really noteworthy? When that is the case then certainly mark as significant in your data presentation, but don’t highlight it in the text.

So to take the example above,

Weight in the intervention group was significantly different to the control group (P<0.01) and there was an interaction between the effect of intervention and sex (P<0.05, Table 2).

We can translate it into human:

As shown in Table 2, at the end of the study the participants in the intervention weighed on average 5kg less than the control group (P<0.01). This effect appeared to be greater in men than in women (interaction of intervention and sex P<0.05) and whilst men in the intervention were on average 5.8 kg lighter than control men the difference was marginal among women (0.5 kg difference).

So by all means, wear your most sophisticated and discerning stats-goggles when you do the analysis, but then put all of the P values to one side and look at the summary data with a fresh eye. Think about what it really shows and how you might explain it and then reinstate the P values. These could even be used in the same way that you use references in the introduction and discussion. They are the evidence base on which you construct your argument.


Having too many tables and figures

It is so tempting to put every piece of data that you have collected into your paper, but rarely necessary. I am not advocating that you hide anything from your reader, but you should leave out material which does not add to the story you are telling, or which is unnecessary in order to show how you have tested your hypotheses. The fact that you collected a piece of data does not necessitate publication. A common problem is over use of Figures, with data that would fit nicely fit into a single Table being split up into many graphs (4 graphs with 4 sub-panels might easily fit into one Table that is just as informative). As a rule of thumb a combined total of 7-8 Tables and Figure should be the maximum that you are considering for presentation in the paper. Most editors would prefer less than this. In the modern age of online publication, there is generally the facility to publish supplementary material which is not included in the paper proper. This is a great way for you to get everything you measured into the public domain, without cluttering the paper.


Presenting raw data

It is very rare in the biological and medical sciences that we would ever present individual data points from a single individual or other sample type. Always use appropriate summaries; mean (or median) with a measure of variance such as standard deviation, standard error of the mean or range.


Forgetting the units of measure

Just don’t! This is something we have drummed into us at school, so why forget it when writing for a scientific journal! No excuses. But… make sure that you are using the correct units for the journal that you are writing for. Most of them will have their own quirks- should you be quoting mg/dL, g/L , mM or mmol/L? It is best to check as you will end up with annoying corrections otherwise.


Repeating presentation of data

Editors frown upon the double presentation of data, maybe showing the same item in a Table and then showing it again in a Figure. This is unnecessary padding and you will be asked to remove it (assuming it doesn’t tip the editorial decision over into the Reject zone).

Early view- a FFQ for iodine in women

Validation of a short food frequency questionnaire specific for iodine in UK females of childbearing age

Combet and Lean. JHND Early View.


Widespread subclinical iodine insufficiency has recently been reported in Europe, based on urinary iodine using World Health Organization/Food and Agriculture Organization criteria, in particular among young women. Although urinary iodine concentration (UIC) is a useful measurement of the iodine status in a population, it does not provide an insight into the habitual iodine intake of this population. This is compounded by the fact that very few iodine-specific food frequency questionnaires (FFQ) have been validated so far. The present study aimed to develop and validate a new, simple, rapid survey tool to assess dietary iodine exposure in females of childbearing age.


Iodine was measured in a duplicate 24-h urine collection. Iodine intake was measured with duplicate 4-day semi-quantitative food diaries and the FFQ. Correlation, cross-classification and Bland–Altman analyses were used to estimate agreement, bias and the reliability of the method. The triangular (triad) method was used to calculate validity coefficients.


Forty-three women, aged 19–49 years, took part in the validation of the 17-items FFQ. Median (interquartile range) UIC was 74 (47–92) μg L−1, which is indicative of mild iodine insufficiency. The FFQ showed good agreement with food diaries with respect to classifying iodine intake (82% of subjects were classified in the same or adjacent quartile). The FFQ was moderately correlated with the food diaries (rs = 0.45, P = 0.002) and urinary excretion in μg L−1 (rs = 0.34, P = 0.025) but not in μg day−1 (P = 0.316). The validity coefficients were 0.69, 0.66 and 0.52 for the food diaries, FFQ and urinary iodine excretion, respectively.


The FFQ provides a rapid and reliable estimate of dietary iodine exposure to identify those population subgroups at risk of iodine deficiency.