¹Department of Physics, University of Caldas, Colombia

²Department of Medical Physics, University of Leeds, Leeds, UK

³Department of Paediatrics, University of Leeds, Leeds, UK

Correspondence to: C H Gonzalez, c/o JA Evans, Department of Medical Physics, Leeds General Infirmary, Gt George Street, Leeds LS1 3EX, UK. E-mail: labie@cumanday.ucaldas.edu.co

Guarantor: CH Gonzalez.

Contributors: JAE, SWS and PH were all involved in study conception, design, interpretation of results and approving final manuscript.

Objectives: (1) To develop a scale that is useful in evaluating the accuracy of multifrequency bioelectrical impedance analysis (MF-BIA) in the assessment of body water volumes against the accepted gold standard measurements based on isotope-dilution and total body potassium (TBK). (2) To perform a pilot test of the scale.

Design: A scale was developed to evaluate the accuracy of MF-BIA in the assessment of body water volumes. Questions were obtained from reading the scientific literature and discussions involving the four authors. Three of these and two additional independent readers pre-tested the scale. A weighting was identified for each question and a pilot test with a sample of 10 articles (different to those used for the questionnaire performance) was conducted. A further validation was carried out with a second set of 20 articles and two additional independent readers.

Results: The kappa statistic expressing the level of agreement between pairs of the first three authors using this scale with 10 articles, was 0.3, 0.4 and 0.6 after the first attempt. A second evaluation after specific changes improved the agreement to 0.8, 0.6 and 0.8. The mean score for 10 articles was 252±36 points from a total score of 400 (63±9%). The evaluation with the second set of 20 articles resulted in a of 0.7 from two pairs of authors. The evaluation with two additional reviewers resulted in a =0.7.

Conclusion: A tool has been developed to assess the accuracy of the MF-BIA technique and to identify methodological components, plan future studies and critically evaluate data in this area. It is likely that this tool may also be used to assess the accuracy of single frequency studies.

Sponsorship: COLCIENCIAS and University of Caldas, Colombia sponsor CH Gonzalez.

European Journal of Clinical Nutrition (2002) 56, 326-337. DOI: 10.1038/sj/ejcn/1601316

bioelectrical impedance; dilution techniques; TBK; body water; scoring system

Introduction

A major challenge in clinical practice is to accurately monitor acute and chronic changes in hydration during various diseases and their clinical management ( Johnson et al, 1996; Pullicino et al, 1992). Thus, the development of methods for measuring and monitoring the volume of body water and its distribution has been of increasing recent interest (Ellis & Wong, 1998). Methods based on bioelectrical impedance analysis (BIA) or its multifrequency variant (MF-BIA) have attracted attention because of their low cost and high convenience. The validation of these techniques normally involves using isotope dilution (Brummer et al, 1992) and total body potassium (TBK) as gold standards (De Lorenzo et al, 1995). There has been a rapid proliferation of BIA literature. A search from 1985 to 1990 with the keyword: 'bioelectrical impedance' (when Lukaski's article appeared), retrieved 73 studies; from 1991 to 1995 there were 271 articles and, from 1996 to 2001, 523 articles. This was only in Medline database. However there are conflicting and confusing results and conclusions about its accuracy and its utility (Schoeller, 1996a). Variability in the quality of BIA studies may have contributed to these inconclusive and conflicting findings. Differences of quality across the studies may indicate that the results of some are more reliable than those of others. Low quality studies are more likely to produce incorrect inferences and conclusions regarding the accuracy and usefulness of BIA. Therefore, evaluation of BIA should be based on evidence from high-quality studies.

The purpose of this paper was to describe a scale to evaluate the quality of BIA studies and to define and discuss the evaluation criteria which might be more generally used.

Scale development methods

The process used to perform our study consisted of eight steps shown in Figure 1. Steps 1-7 are preparation with the final evaluation (step 8) following at the end. The details of steps 1-7 are described below.

1. Literature review to identify key characteristics to be extracted from each paper through examination of

1. Key characteristics identified as being paramount in published checklists evaluating quality of studies published in the health technology assessment literature to identify general headings for evaluation. (Bland et al, 1985; Cho & Bero, 1994; Detsky et al, 1992; Khan et al, 1996; Moher et al, 1995; 1996; NHS Centre for Reviews and Dissemination, 1996; Oxman, 1994)
2. Possible bias (sampling, diagnostic access, spectrum, co-intervention, verification, withdrawal, work-up, incorporation disease progression, observer variability, diagnostic review, test review, clinical review and loss to follow-up bias (Kelly et al, 1997)
3. Articles about BIA, TBK and dilution techniques standardisation (Deurenberg et al, 1988; Kushner et al, 1996; National Institutes of Health Technology Assessment Conference Statement, 1994; Schoeller et al, 1985; Schoeller, 1996b; Lykken et al, 1983)
4. Articles about the use of statistics in scientific literature (Altman, 1980, 1991; Altman et al, 1983; Emerson & Colditz, 1983; Gardner et al, 1983; 1986; Jenicek, 1989)
5. Views expressed by the authors who have some knowledge and experience in the area: two medical physicists, one paediatrician and one physician

This process resulted in steps 2 and 3 below

2. Definition of quality

Quality as defined for our purposes of evaluating BIA estimates of fluid volumes was based on the following subheadings:

1. Study design
2. Bias
3. Methodological performance
4. Statistical analysis

3. Development of a questionnaire for specific BIA evaluation

Under the headings identified above, a questionnaire was composed consisting of 66 questions. Additionally a form was developed to extract the data from the articles.

4. Evaluation of the questionnaire by iterative development using three independent observers

Three authors with different backgrounds (two medical physicists and one physician studying the BIA technique) pre-tested the questionnaire three times using 10 articles in order to decide which questions should remain in the final questionnaire. Specific suggestions were:

1. To record a 'No' when a procedure was specifically mentioned as not done but 'Not clear' if the procedure was mentioned but without sufficient detail to allow replication
2. To split compound questions in order to achieve more specificity
3. To eliminate duplication in the questions

The final questionnaire consists of 21 explicitly defined questions (Appendix part 1, part 2, part 3 and part 4). Questions 12 and 13 refer to three sub-questionnaires about procedures on BIA, dilution and TBK techniques respectively. Possible responses are: 'Yes', 'No', 'Not clear', or 'Not applicable'.

The questionnaire was divided into four main sections:

1. Section A examines the design of the study and inclusion/exclusion criteria for its subjects
2. Section B looks for description of the subjects and how they were selected
3. Section C is related to the procedures for carrying out the BIA and the gold standard techniques
4. Section D concerns calculations and statistical analysis

Forty-two percent of questions are related only to the quality of the research design, 16% to the quality of the report writing and structure and 42% consider both topics. The original version of the questionnaire was modified to make it simpler and more self-explanatory.

5. Identify weighting for each characteristic

A theoretical weighting scheme was used to score the studies according to Streiner and Norman (1995). In this case the weighting scheme was established from an analysis of the literature and from local experimentation. The following factors were given heavy weighting: study design; sample selection; BIA; TBK and dilution procedures, hydration status; and statistical analysis. Each section was scored on a continuous scale from 0 to 100 for positive answers. Within a section each item was weighted according to its impact on results when data were available in the scientific literature or when a variable could influence the results in more than one technique (references used to give the weight for each variables are at www.medphysics.leeds.ac.uk/chgc/vienaref.htm). When the data were not available in the literature, questions were weighted equally. Negative answers did not receive a score and the 'Not applicable' option allowed reviewers not to penalise a paper if one item was not necessary for the study. The maximum possible score was 400 points. Figure 2 shows the extent to which each item contributes to the total score for an article.

6. Pilot test

When the reviewers had agreed on the final version of the questionnaire, three evaluations were conducted for the assessment of reproducibility and reliability. An additional test was performed to evaluate agreement in data extraction. For the first evaluation a computerised MEDLINE search was performed to identify articles which used BIA and dilution or TBK techniques and which were published between 1969 and 1998 (inclusive). This covered the period between the original article by Hoffer et al (1969) and the time of the study. The exercise was performed as if these articles were to enter to a meta-analysis. The search strategy was exploration by bioelectrical impedance analysis (or 17 substitutes) and body water (or 13 substitutes) or dilution (or nine substitutes) and TBK from 1969 to 1998. The search was limited to studies with human beings, English and Spanish languages. Letters, abstracts and editorials were also excluded. The initial search yielded 161 citations.

One of the reviewers read the abstracts and selected those that had used dilution techniques (TBW/ECW) and/or total body potassium (ICW) as gold standard. Seventy-two studies were eligible. From these, random numbers generated by a calculator selected five articles and five additional articles were chosen for the pilot test to give a variety of ages and clinical conditions.

Three reviewers tested the final questionnaire. The revision was made on the basis of inter-rater agreement and reviewers' comments on content validity of individual questions. The reviewers worked independently. To minimise observer bias, they were given masked photocopies of the 10 articles. The methods sections of these studies were photocopied in such a way that the author and journal names, dates, and all other reference information had been omitted. General guidelines were given on the cover sheet for the use of the scale. An additional questionnaire to describe the areas of particular strength, reason for doubt or for rejection, time spent with the evaluation of each study, and the influence of the knowledge of one of the articles on the final decision on it was attached to each paper.

The first evaluation of reproducibility and reliability produced results which were poor, as expressed by the kappa statistic (<0.6). A second evaluation was performed after making the following changes:

1. The low agreement in the first evaluation was generally the result of discrepancies in the use of the response 'Not applicable' and probably fast reading of the articles. Therefore, the results of the questionnaires were corrected when the answers were unambiguous (side of the body on which the measurements were performed, baseline samples for dilution techniques, etc)
2. The instruction manual was expanded to explain in more detail definitions of terms and the importance of each question. At this time, the reviewers were given a six-page set of written instructions
3. Reviewers suggested that when references were cited in the 'Methods' section, these references should be included with the article. For the second evaluation the relevant bibliography was therefore attached to each article

The kappa statistic for the same 10 articles was improved after these changes (=0.6). An evaluation was done with a form developed for data extraction. Three reviewers tested the form by using three of the 10 articles (results not shown).

Selection of the articles: The acceptability of each article was described as 'Definitely yes', 'Probably yes' or 'Not acceptable' on the basis of the score. Differences among reviewers were resolved by re-examining the report and discussing the particular point.

Inclusion criteria: Studies were selected after three assessments by unanimous decision or by agreement between two of the three reviewers: studies with a score 60% of the total possible score was qualified as 'Definitely yes'; studies that were scored from 50-59% of the total possible score were qualified as 'Probably yes'; and the final inclusion or exclusion of these articles was made by a third evaluation between the readers.

Exclusion criteria: Studies with a score of <50% of total possible score or studies classified as 'Probably yes' which after a third evaluation still had an score <60%.

7. Further scale validation

To test whether the repeated review of the same 10 articles had influenced the results, a second different set of 10 masked articles was reviewed independently by two pairs of the of the reviewers (20 different articles in total).

After a consistent agreement was obtained between us, two independent reviewers with experience in the bioimpedance area but not involved in the development of the instrument were asked to qualify three randomly selected and masked articles.

How to use the scale: After revision of the study, each question was answered by ticking the appropriate box in the form. Only questions with positive or 'Not applicable' answers were weighted. When the study compared BIA with both dilution and TBK techniques, the weights applied to these questions were added to the gold standard score.

Statistical analysis

Percentage agreement and kappa statistic were used for the analysis of agreement between reviewers on study choice. Acceptable agreement was set at a kappa level from 0.6 to 1.0. Kappa interpretation was made according to Altman (1991). The score of the articles is expressed as a percentage of the total possible score.

Results

Three reviewers who used the quantitative scale to rank independently the set of 10 articles evaluated reproducibility, reliability and applicability. There were only three instances of missing data: two from one reader and one for another reader. Tables 1 part 1, part 2, 1a, 1b, and 1c shows the final questionnaire with the references on which the score was based and Appendix part 1, part 2, part 3 and part 4 the forms used for each article. The test required approximately 20 min per article per reviewer. Table 2 shows the kappa values and percentage inter-rater agreement between reviewer pairs for their summary ranking scores on the same 10 articles in two occasions. The final result of the evaluation of the articles is given in Table 3. after the 'Probably yes' articles were reviewed again among the three reviewers. The mean of reviewer scores for 10 articles was 251.6±35.9 points from a total score of 400 (62.9±8.98%; Table 4). In the further validation of the scale between two pairs of reviewers with another set of 10 articles per pair, the kappa statistic was 0.7 (CI 0.6-0.7) and 0.7 (CI 0.6-1.3).

The agreement between the two independent reviewers was =0.7 (confidence interval 0.6-0.8) without corrections for solving differences. Altman (1991) considers this agreement as good. The average time for new users of the scale (including the manual) was 17 min.

Discussion

In order to perform a systematic review of primary research of variable quality, it is necessary to have an adequate assessment of study quality which guarantees that the results are likely to be free from bias and consequently can be generalisable (Khan et al, 1996).

We developed a standardised scale that could be used to score the quality of BIA studies in comparison with gold standard techniques. The scale can be used as a simple checklist or to provide a quantitative summary score. It does not apply uniform weighting across all questions. It weights them according to defined criteria. Scoring for each attribute thereby defines its relative contribution to the total score.

We chose '>60%' as the standard for the highest category of acceptance to perform the pilot test. It was similar to Chalmers et al (1990) in which a score of >6/9 was regarded as good. This was a cohort study of summary reports of controlled trials. In another study of 242 randomised controlled trials Rochon et al (1994) found an average score of 37.2±13.1%. We concluded that the greater of these two criteria was more appropriate.

This scale is an innovation in BIA studies and it is important for several reasons: firstly, because reviewers can use it to identify critical methodological components in this area. Secondly, researchers can use it to plan their studies; using the scale ensures that important factors such as room temperature, which can strongly affect results, will not be forgotten. Thirdly, literature reviewers wishing to summarise information on BIA can use it to evaluate and rank the quality of existing pertinent literature. The scale can help authors and editors to ensure that all the fundamental information needed in the reports is included (Lionel & Herxheimer, 1970).

The scale may be also used as a tool for standardisation of studies that could be candidates for inclusion in metaanalyses. Meta-analysis is a novel scientific technique, which allows a more objective appraisal of the evidence than do traditional narrative reviews. It provides a more precise estimate of the effectiveness of health technology and may explain variability between the results of individual studies (Hardy & Thompson, 1998).

Methodological quality of a study is defined as a minimisation of systematic bias and consistency of conclusions with results (Cho & Bero, 1984). It is possible to determine the methodological quality of a study only to the extent that the study design and analytical methods are reported. Our process measures criteria critical to achieving high methodological quality; however, in consequence, these criteria should be adequately reported in any article using BIA, dilution and TBK techniques. The lack of important items of description in some apparently well-executed studies may at times be due to the short space authorised by editors for publication.

The ideal kappa for agreement between raters is +1 (Altman, 1991). In the pilot test, the first agreement between the reviewers was below the desirable score (>0.6). Similar studies report a kappa average greater than 0.7 for all reviewers (Khan et al, 1996). A second assessment of the same 10 articles was made using more detailed instructions and more precise definitions of terms. The kappa for this assessment was within the desirable range. Finally, further discussions and clarification resulted in complete agreement (Table 3).

The scale had good reliability and validity when used by reviewers from different backgrounds who worked completely independently and without lengthy training sessions. Independence of reviewers avoids the systematic bias that can be introduced by consensus conferencing if one of the reviewers has convincing opinions (Cho & Bero, 1994).

We performed a further evaluation of the scale by using a second different set of 20 masked articles the agreement was 'good' according to the Altman (1991) classification. Two additional reviewers that work in the bioimpedance field obtained also a 'good' agreement without any further correction. This reinforces the reliability of the scale.

Any method of evaluating quality of published scientific studies stimulates discussion of validity. The present scale has evolved over a period of 3 y and will probably be changed as experience is gained in this field. Among the possible defects are variation due to subjective judgements, inaccuracies due to incomplete reporting, inclusion of questions that might not be considered meritorious by some experts, and exclusion of some questions considered important by others. To some extent, these problems have been handled by the method of differential photocopies of the studies and by multiple scoring followed by discussion.

We are aware of the influence of the machine in the MF-BIA results. However, we have not made a judgement at this initial stage as to which machine is the most accurate and therefore the identification of the machine does not properly form a part of the quality assessment instrument. Having developed the tool it would then be possible to use it to study the influence of machine type on the accuracy of measurements and this would indeed be a useful, but separate study.

The issue of algorithm selection (regression, mixture theory, etc) is slightly different and we believe that explicit statement of the derived algorithm used to calculate body water is a characteristic of a good quality study. Therefore, in question 19 we address the matter related to the algorithm. The judgement of whether the algorithm used was the correct one or not is beyond the study objectives. In this study, we did not set out to say which one was better, although a separate study using this tool could tackle this issue.

To develop the scale we have made and extended review of literature. However, we would like to discuss with others the performance of the instrument and we are ready to update it when it may be necessary. We will prepare the scale to be discussed in a web page in the near future (URL:http://www.medphysics.leeds.ac.uk/~chgc/BIAscoresystem.htm).

This scale is intended for use in a future evaluation of BIA literature compared with dilution and TBK techniques.

Altman DG. (1980). Statistics and ethics in medical research. Br. Med. J., 28: 1336-1338.

Altman DG. (1991). Practical Statistics for Medical Research. London: Chapman and Hall.

Altman DG, Gore SM, Gardner MH, Pocock SJ. (1983). Statistical guidelines for contributors to medical journals. Br. Med. J., 286: 1489-1493.

Armstrong LE, Kenefick RW, Castellani JW, Riebe D, Kavouras SA, Kuznicki JT. (1997). Bioelectrical impedance spectroscopy technique: intra-, extracellular and total body water. Med. Sci. Sports Exerc., 29: 1657-1663. MEDLINE

Baar J, Tannock I. (1989). Analysing the same data in two ways: a demonstration model to illustrate the reporting and misreporting of clinical trials. J. Clin. Oncol., 7: 969-978. MEDLINE

Bailar JC, Mosteller F. (1988). Guidelines for statistical reporting in articles for medical journals. Ann. Intern. Med., 108: 266-273. MEDLINE

Bland JM, Altman DG. (1986). Statistical methods for assessing agreement between two methods of clinical measurement. Lancet, 8: 307-310.

Bland JM, Jones R, Bennet S, Cook DG, Haines AP, Macfarlane AJ. (1985). Is the clinical trial evidence about new drugs statistically adequate? Br. J. Clin. Pharmac., 19: 155-160.

Brummer RJM, Bengtsson BA, Bosaeus I. (1992). Validation of body composition determination by bioelectrical impedance analysis in acromegaly. Eur. J. Clin. Nutr., 46: 47-52. MEDLINE

Chalmers TC, Smith H, Blackburn B, Silverman B, Schroeder B, Reitman D, Ambroz A. (1981). A method for assessing the quality of a randomised control trial. Control Clin. Trials, 2: 31-49. MEDLINE

Chalmers I, Adams M, Dckersin K, Hetherington J, Tarnow-Moedi Q, Meinert C, Tonascia S, Chalmers TC. (1990). A cohort study of summary reports of controlled trials. J.A.M.A., 263: 1401-1405. MEDLINE

Cho MK, Bero LA. (1994). Instruments for assessing the quality of drug studies published in the medical literature. J.A.M.A., 272: 101-104. MEDLINE

Cornish BH, Ward LC, Thomas, BJ, Jebb SA, Elia M. (1996). Evaluation of multiple frequency-bioelectrical impedance and Cole-Cole analysis for the assessment of water volumes in healthy humans. Eur. J. Clin. Nutr., 50: 159-154. MEDLINE

Cox R, Soeters PB. (2000). Validation of bio-impedance spectroscopy: effects of degree of obesity and ways of calculating volumes from measured resistance values. Int. J. Obes. Relat. Metab. Disord., 24: 271-280. MEDLINE

De Lorenzo A, Candeloro N, Andreoli A, Deurenberg P. (1995). Determination of intracellular water by multifrequency bioelectrical impedance. Ann. Nutr. Metab., 39: 177-184. MEDLINE

Detsky AS, Naylor DC, O'Rourke K, McGeer AJ, L'Abbe KA. (1992). Incorporating variations in the quality of individual randomised trials into meta-analysis. J. Clin. Epidemiol., 45: 255-265. MEDLINE

Deurenberg P. (1996). Multifrequency-bioelectrical impedance a comparison between the Cole-Cole modelling and Hanai equations with the classical impedance index approach. Ann. Hum. Biol., 23: 31-40. MEDLINE

Deurenberg P, Westrate I, Paymans I, Koy K. (1988). Factors affecting bioelectrical impedance measurements in humans. Eur. J. Clin. Nutr., 42: 1017-1022. MEDLINE

Ellis KJ. (1996). Whole-body counting and neutron activation analysis. In Human Body Composition, eds. AF Roche, SB Heymsfield & TG Lohman. 45-61, Champaign, IL: Human Kinetics.

Ellis KJ, Wong WW. (1998). Human hydrometry: comparison of multifrequency bioelectrical impedance with ²H₂O and bromide dilution. J. Appl. Physiol., 85: 1056-1062. MEDLINE

Emerson JD, Colditz GA. (1983). Use of statistical analysis in The New England Journal of Medicine. New Engl. J. Med., 309: 709-713. MEDLINE

Evans M, Pollock AV. (1985). A score system for evaluating random control clinical trials of prophylaxis of abdominal surgical wound infection. Br. J. Surg., 72: 256-260. MEDLINE

Forbes GB. (1987). Techniques for estimating body composition. In Human Body Composition. 5-45, New York: Springer.

Gardner MJ, Altman DG, Jones DR, Machin D. (1983). Is the statistical assessment of papers submitted to the British Medical Journal effective? Br. Med. J. Clin. Res., 286: 1485-1488.

Gardner MJ, Machin D, Campbell MJ. (1986). Use of checklists in assessing the statistical content of medical studies. Br. Med. J., 292: 810-812.

Gonzalez C, Evans JA, Smye SW, Holland P. (1999). Variables affecting Bioimpedance analysis measurements of body water. Med. Biol. Engng. Comp., 37: 106-107.

Goodman SN, Berlin J, Fletcher SW, Fletcher RH. (1994). Manuscript quality before and after peer review and editing at Annals of Internal Medicine. Ann. Intern. Med., 121: 11-21. MEDLINE

Hannan WJ, Cowen SJ, Plester CE, Fearon KVH, Debeau A. (1995). Comparison of bioelectrical impedance spectroscopy and multifrequency bioimpedance analysis for the assessment of extracellular and total body water in surgical patients. Clin. Sci., 89: 651-658. MEDLINE

Hardy RJ, Thompson SG. (1998). Detecting and describing heterogeneity in meta-analysis. Stat. Med., 17: 841-856. MEDLINE

Ho TL, Kushner RF, Schoeller DA, Gudivaka R, Spiegel DM. (1994). Bioelectrical impedance analysis of total body water in haemodialysis patients. Kidney Inte., 46: 1438-1442.

Hoffer EC, Meador CK, Simpson DC. (1969). Correlation of whole-body impedance with total body water volume. J. Appl. Physiol., 27: 531-534. MEDLINE

Jebb SA, Elia M. (1993). Techniques for the measurement of body composition: a practical guide. Int. J. Obes. Relat. Metab. Disord., 17: 611-621. MEDLINE

Jenicek M. (1989). Meta-analysis in medicine. Where we are and where we want to go? J. Clin. Epidemiol., 42: 35-44. MEDLINE

Johnson DW, Thomas BJ, Fleming SJ, Westhuyzen J, Moran D, Ward LC. (1996). Monitoring of extracellular and total body water during haemodialysis using multifrequency bioelectrical impedance analysis. Kidney Blood Press Res., 19: 94-99. MEDLINE

Kelly S, Berry E, Proderick P, Harris KM, Cullingworth J, Gathercole L, Hutton J, Smith MA. (1997). The identification of bias in studies of the diagnostic performance of imaging modalities. Br. J. Radiol., 70: 1028-1035. MEDLINE

Khan KS, Daya S, Jadad A. (1996). The importance of quality of primary studies in producing unbiased systematic reviews. Arch. Intern. Med., 156: 661-666. MEDLINE

Kushner RF, Gudivaka R, Schoeller DA. (1996). Clinical characteristics influencing bioelectrical impedance analysis measurements. Am. J. Clin. Nutr., 64: 423-427.

Lionel NDW, Herxheimer A. (1970). Assessing reports of therapeutic trials. Br. Med. J., 3: 637-640. MEDLINE

Lykken GI, Lukaski HC, Bolonchuck WW, Sandstead HH. (1983). Potential errors in body composition as estimated by whole body scintillation counting. J. Lab. Clin. Med., 4: 651-658.

McMaster University Health Sciences Centre. (1981). How to read clinical journals: III. To learn the clinical course and prognosis of disease. Can. Med. Assoc. J., 124: 869-872. MEDLINE

Moher D, Jadad AR, Nichol G, Penman M, Tugwell P, Walsh S. (1995). Assessing the quality of randomised controlled trials: an annotated bibliography of scales and checklists. Control Clin. Trials., 16: 62-73. MEDLINE

Moher D, Jadad AR, Tugwell P. (1996). Assessing the quality of randomised controlled trials. Int. J. Tech. Assess. Health Care., 12: 195-208.

Mulrow CD, Linn WD, Gaul MK, Pugh JA. (1989). Assessing quality of a diagnostic test evaluation. J. Gen. Intern. Med., 4: 288-295. MEDLINE

National Institutes of Health Technology Assessment Conference Statement. (1994). Bioelectrical impedance analysis in body composition. Am. J. Clin. Nutr., 64: 524S-532S.

NHS Centre for Reviews and Dissemination. (1996). Undertaking Systematic Reviews of Research on Effectiveness. CRD Guidelines for those Carrying Out or Commissioning Reviews (CRD report no. 4). York: NHS Centre for Reviews and Dissemination, University of York.

Oxman AD. (1994). Checklists for review articles. Br. Med. J., 309: 648-651.

Pullicino E, Coward WA, Elia M. (1992). The potential use of dual frequency bioelectrical impedance in predicting the distribution of total body water in health and disease. Clin. Nutr., 11: 69-74.

Reisch JS, Tyson JE, Mize G. (1989). Aid to the evaluation of therapeutic studies. Paediatrics, 84: 815-827.

Rochon PA, Gurwitz JH, Cheung CM, Hayes JA, Chalmers TC. (1994). Evaluating the quality of articles published in journal supplements compared with the quality of those published in the parent journal. J.A.M.A., 272: 108-113. MEDLINE

Schoeller DA. (1996a). Bioelectrical impedance analysis for the measurement of human body composition: where do we stand and what is the next step? Nutrition, 12: 760-762.

Schoeller DA. (1996b). Hydrometry. In Human Body Composition, eds. AF Roche, SB Heymsfield & TG Lohman. 25-44, Champaign, IL: Human Kinetics.

Schoeller DA, Kushner RF, Taylor P, Dietz WH, Bandini L. (1985). Measurement of total body water: isotope dilutions techniques. In Body Composition Assessment in Youth and Adults: Sixth Ross Conference on Medical Research, ed. AF Roche. 124-129, Columbus, OH: Ross Laboratories.

Streiner DL, Norman GR. (1995). From items to scales. In Health Measurement Scales: a Practical Guide to their Development and Use. 85-103, Oxford: Oxford University Press.

Van Marken Lichtenbelt WD, Snel YEM, Brummer RJM, Koppeschaar HPF. (1997). Deuterium and bromide dilution and bioelectrical impedance spectrometry independently show that growth hormone-deficient adults have an enlarged extracellular water compartment related to intracellular water. J. Clin. Endocrinol. Metab., 82: 907-911. MEDLINE

Visser M, Deurenberg P, Staveren WA. (1995). Multi-frequency bioelectrical impedance for assessing total body water and extracellular water in elderly subjects. Eur. J. Clin. Nutr., 49: 256-266. MEDLINE

Westerterp KR, Wouters L, Van Marken Lichtenbelt WD. (1995). The Maastricht protocol for the measurement of body composition and energy expenditure with labeled water. Obes. Res., 3: 49-57. MEDLINE

Working Group on Recommendations for reporting of clinical Trials in the Biomedical Literature. (1994). Call for comments on a proposal to improve reporting of clinical trials in the biomedical literature. Ann. Intern. Med., 121: 894-895. MEDLINE

Appendix 1

Figure 1 Scale development process.

Figure 2 Contribution (%) of characteristics to score of MF-BIA/dilution studies.

Table 1 Scale to evaluate studies which use bioelectrical impedance analysis compared to dilution and TBK techniques in estimation of body water volumes

Table 1a Question no. 12 of the questionnaire to evaluate studies BIA procedures

Table 1b Question no. 13 dilution techniques standardisation

Table 1c Question no. 13 total body potassium standardisation

Table 2 Agreement between three reviewers

Table 3 Selection of articles

Table 4 Final score for 10 articles

Appendix Test to evaluate BIA studies

Appendix BIA technique standardisation

Appendix Dilution technique standardisation

Appendix TBK standardisation