Evaluation of validity of British anthropometric reference data for assessing nutritional state of elderly people in Edinburgh: cross sectional study

Introduction

Malnutrition is common, serious, and largely unrecognised in British hospitals.1 It is also particularly common in old age.2 3 Good anthropometric reference data are therefore fundamental in assessing the nutritional state of elderly people as well as in studies quantifying the prevalence of malnutrition and screening for malnourished people. However, there are few anthropometric data from large representative samples of elderly people,4 and geographical variation in anthropometric variables might be large.5

Two sets of anthropometric reference data are widely used in nutritional assessment of elderly people in Britain. The first set is data from Nottingham on weight, demispan, demiquet (weight/demispan2) for men and mindex (weight/demispan) for women.6 These data were obtained in the 1980s from a representative sample of 890 people aged 65 or over who were not living in institutions and have been suggested for use to identify people at extremes of the distribution.7 The second set comprises data on body mass index, mid-upper arm circumference, triceps skinfold thickness, and arm muscle circumference collected from a broadly representative sample of about 1500 elderly subjects living in South Wales in the 1970s (7% in institutions).8 We conducted a study to evaluate the appropriateness of these anthropometric data in nutritional assessment of a representative sample of elderly people living in Edinburgh.

Subjects and methods

We randomly selected a sample of men and women aged 75 or over who were living independently and registered with two general practices in Edinburgh. We used a quasi-random sampling frame in which we contacted every nth eligible patient from the age-sex registers. The patients were taking part in a larger study of potential risk factors for poor nutritional state in people over 75. For that study it was calculated that if the true correlation between any two variables was 0.2, a sample size of 200 would give an 80% chance of detecting the association as significant at the 95% significance level (P=0.05). Assuming a 35% non-response rate, about 300 people would need to be asked to join the study.

We sent out a total of 246 contact letters to potential subjects. Two hundred people were recruited to the main study, giving a response rate of 81%. Seventeen patients declined to take part in the study, including having a terminal illness (n= 4), looking after an ill partner (n=3), or being in the process of moving into a nursing home, long stay ward, or residential accommodation (n=10); 23 people withheld their consent and six did not reply to the contact letter.

All measurements were performed in the subject's own home in an attempt to improve the recruitment and participation rates. Written informed consent was obtained before measurements were made, and the study was approved by Lothian Health Board ethics committee.

Each subject was weighed on calibrated portable scales to the nearest 0.5 kg without shoes but in indoor clothing, which was then accounted for. Any extreme signs of oedema were noted, and any subjects affected were excluded from weight calculations.

The simplest form of nutritional assessment is where weight is adjusted for height. However, because of the problems of spinal curvature and kyphosis that occur with advancing age, we used two standard alternative approaches for estimating height—demispan and knee height. Demispan is the distance from the web between the third and fourth fingers along the outstretched arm to the sternal notch with the arm in the corneal plane. It was measured on the left arm unless it had been affected by disease or disability by using a plastic tape measure which had a button attached to anchor it at the base of the subject's fingers.9 Measurements were made to the nearest 10 mm. Knee height measurements were made with the Ross knee height calliper, which has been shown to have acceptable accuracy and reliability.10 The left leg was measured with the subject supine unless it had been affected by disease or disability.11 Measurements were made to the nearest 1 mm, and body height was estimated by using the equations produced by Chumlea et al.11

We calculated body mass index (kg/m²) using the stature estimated from the knee height measurement. This procedure produces unbiased estimates of stature for groups.12 We also calculated mindex (weight/demispan (kg/m)) for women and demiquet (weight/demispan²(kg/m²)) for men.6

Measurements of the upper arm were made on the left arm unless it had been affected by disease or disability. We measured mid-upper arm circumference to the nearest 1 mm with a plastic tape measure and triceps skinfold thickness with calibrated Holtain skinfold callipers to the nearest 0.2 mm using standard techniques.13 Arm muscle circumference and corrected arm muscle area were calculated from these measurements14: arm muscle circumference (cm)=mid-upper arm circumference (cm)-(triceps skinfold thickness (mm)xπ). Corrected arm muscle area (cm²)=(arm muscle circumference (cm)²/4π)-10 (for men) or (arm muscle circumference(cm)²/4π)-6.5 (for women).

Additional social information, including the subject's postcode, was collected to determine the extent to which the sample was representative of the population from which it was drawn. Analysis was carried out using the Scottish ACORN (a classification of residential neighbourhoods) classification system, which is formulated from information on over 100 variables collected from the 1991 census data. Socioeconomic classification in the system is based on factors such as home ownership, age, health, employment, and occupation as well as special factors such as floor of residence and overcrowding. The classification is divided into 43 types that make up eight groups. Because of our relatively small sample size we restricted the classification to the group level.

Two sample unpaired t tests were used to investigate any significant difference between the Edinburgh data and those in the reference data from Nottingham6 and South Wales.8 Standard deviations were not available for the South Wales data and we therefore used the Edinburgh standard deviations in the analysis. The Kolmogorov-Smirnov goodness of fit test was used to ascertain that the data were normally distributed for each parameter.

Results

The sample was broadly representative of Edinburgh in terms of socioeconomic characteristics (fig 1). The mean (SD) age of non-participants (81.6 (4.9) years, range 75-95) was similar to the age of those who did participate (80.4 (4.9) years). The ratio of men to women was 62:138.

Fig 1

Socioeconomic classification by ACORN system of 200 people aged 75 years or over who were living independently in Edinburgh compared with classification for weighted population base and for total populations of Edinburgh and Scotland

• Download figure
• Open in new tab
• Download powerpoint

We obtained weight measurements from 188 (94%) subjects (59 men, 129 women), demispan and knee height from 189 (95%; 59 men, 130 women), and upper arm anthropometric data from 185 (92.5%; 59 men, 126 women). The distribution of each parameter was not significantly different from the normal distribution.

Men and women in Edinburgh were significantly shorter than those in Nottingham (0.79 m v 0.72 m, P<0.05 and 0.8 m v 0.73 m, P<0.01 respectively). In addition, table 1) shows that demiquet was significantly greater among men in Edinburgh than Nottingham (P<0.01).

Mid-upper arm circumference, triceps skinfold thickness, and arm muscle circumference were significantly greater in Edinburgh than in Wales for men and women of all ages (table 2). Body mass index was not significantly different for women in the two groups but was significantly greater in Edinburgh men aged 75-79 years (P<0.001) and 80-84 years (P<0.01) than for men in Nottingham.

Discussion

This study highlights significant differences in several important anthropometric indices of nutritional state between three large, broadly representative samples of elderly people from different regions in Britain sampled at different times. These differences would be clinically important when deciding if an elderly person was malnourished and have a substantial impact on outcomes of screening for malnourished elderly patients or on estimates of the prevalence of malnutrition in research and clinical audit.

If the fifth (or even 10th) centile for mid-upper arm circumference, triceps skinfold thickness, or arm muscle circumference from the South Wales data were used as cut offs for malnutrition (as is common1 3) in our Edinburgh sample no men or women would fall below them. No Edinburgh men aged 80 years or over had a body mass index below the 10th centile. For arm muscle circumference and mid-upper arm circumference no men aged 75-9 were below the 25th centile and no older men were below the 50th centile. Furthermore none of our subjects had a corrected arm muscle area below the suggested cut offs of 16.0 cm² for men and 16.9 cm² for women. This measure is widely used as a simple index of undernutrition in old age.14

Comparison with the South Wales data suggests that no one in our population was undernourished. Although differences may exist in the prevalence of undernutrition in elderly populations throughout Britain, it seems unlikely that such large discrepancies can be explained by this. The differences are more convincingly explained by a shift in distribution of anthropometric data in Edinburgh compared with current British reference data.

Similar discrepancies with the current anthropometric reference data have been described for infants15 16 and also in younger adults.17 Furthermore, the difference observed in skeletal size between people in Edinburgh and Nottingham, as shown by comparison of demispan measurements, is supported by data from a national survey of British adults.18 French anthropometric data are very different from British data, and substantial geographical variation has been suggested to occur within France.5

In conclusion, simple anthropometric measurements can provide practical and valid indices of nutritional status.1 19 It is important to assess nutritional state, especially in elderly patients because they are at high risk of malnutrition.20 However, our study suggests that existing reference data for anthropometric nutritional assessment of elderly people in Britain are not representative of all populations. The discrepancy may be due to geographical differences or secular trends. Urgent steps must be taken to establish the reason for the differences and to obtain contemporary reference data appropriate for the whole of Britain. This will ensure that people with malnutrition are reliably identified and enable early and appropriate intervention.

It may be preferable to use alternative criteria to assess nutritional state. One such criterion would be percentage body fat, with minimum and maximum values based on biological or clinical evidence.21 However, developing such an approach will take time, and nutritional assessment is likely to rely on anthropometric data for the foreseeable future. Our results show the need to consider the limitations of reference data when carrying out nutritional assessments.