Overweight and obesity (excess adiposity) are associated with an increased risk of diabetes, arthritis, cardiovascular disease, and certain cancers [1–3]. Over the past twenty years the prevalence of overweight and obesity has risen sharply in many countries [4–7]; and cheap, straightforward techniques for population surveillance of adiposity remain critical. In large epidemiological studies measuring adiposity almost always relies on proxy measures such as the body mass index (BMI) [8].

There is evidence, that proxy measures of central adiposity – waist circumference (WC) and waist to hip ratio (WHR) – are better predictors of adverse health events, including mortality, than BMI [9–11]. A tape measure is also an easier piece of equipment to carry into the field than a stadiometer and a set of weighing scales. Notwithstanding these advantages BMI remains the adiposity metric of choice in most medical research. In 2011, for instance there were 13,909 papers listed in PubMed related to BMI and less than one fifth as many papers related to WC or WHR (n = 2,422).

One factor affecting the adoption (or lack of adoption) of WC or WHR may be the current reliance on inexpensive self-reported BMI measures, and the concomitant uncertainty among the research community about the validity of self measured waist and hip data. Self-reported height and weight, which is used to estimate BMI, has been widely used e.g., [12, 13]; and the validity and biases associated with self-reported height and weight have been the subject of considerable research since at least the 1980s [14]. These have included, for instance, validity studies from Asia [15–17], North America [18], Central America [19], and Europe [20].

In contrast there are relatively few validity studies of self measurement associated with WC and WHR, and with one exception [15], the studies all appear to have come from Europe and North America [21–31]. The lack of validation studies on self measurement of WC and WHR from different populations with different anthropometry, cultural practices, and levels of education will necessarily affect the adoption of WC and WHR as alternative measures of adiposity in medical research. The generalisability of the studies are further affected by variations in the choice of tape measure including paper [22, 30, 31], cloth or plastic [21], marked or unmarked [25], or constant tension tape measures [28]; differences in the device used by participants and technicians [21]; the instructions provided; the presence or absence of light clothing; and the sampling frame (clinical, community, occupational, random, convenience, and so forth).

In this research, we contrasted self measurement and assisted measurement of waist and hip circumference and WHR in a community-based sample from a district town in peninsular Malaysia, using a now standard protocol [32]. We also examined the validity of the measures for scientific research; specifically we examined the extent to which the measurement approach affected statistical relationships with BMI, diastolic and systolic blood pressure, and non-fasting blood glucose. While the need for validation is critical, the research was relevant also to explore the feasibility of self measurement within a context where the permissibility of data collectors taking these measurements is constrained by cultural and social mores.

The minimum, median, and mean WC values for self and assisted measurements were within 1 cm of each other – the maximum was within 1.5 cms. The summary statistics for hip measurement were similarly close. Summary statistics for waist to hip ratios also showed little divergence.

Bland–Altman plots illustrate the difference between two measures against the average value of the two measures [33]. 'Agreement’ is related to both the mean difference between the self and assisted readings, and the amount of variation in the differences. Figure 1 shows the Bland Altman plots for the self and assisted measures of waist and hip circumference and the derived measure of waist to hip ratio. The solid horizontal line indicates the mean difference between the measures. The dashed horizontal lines show the 95% limits of agreement around the mean difference.

Most of the points lie comfortably within the 95% limits of agreement. For the plot of hip measurement, however, there appear to be some extreme outliers showing differences between the assisted measurement and the self-measurement in excess of 10 cms. The data were rechecked for obvious recording errors, but none were apparent. The differences could not be explained by a systematic difference between self- and assisted-measures. In two cases the assisted measure was more than 10 cm greater than the self measurement, and in the other case the self measurement was more than 10 cm greater than the assisted measurement.

The mean differences between the self and assisted measures of waist (-0.81 cm) and hip (-1 cm) circumference indicated a bias, with self-measurement on average lower than assisted measurement. In both cases the mean of the differences was significantly different from 0 (waist: t = -3.22, df = 124, p = .002; hip: t = -4.00, df = 124, p = 0.0001). The 95% limits of agreement – the variation in the differences – for the waist (-0.87 ±6.02 cms ) and hip (-1.1 ±5.98 cms ) were around ±6 cms, with the 68% limits of agreement around ±3 cms. The mean of the differences in the WHR showed a slight, non-significant bias (mean difference = .001, p = .76). Most of the points lie comfortably within the 95% limits of agreement, although there was one point that lay well below the lower limit and a couple of points lay well above the upper limit. The concordance correlation coefficient for the waist, hip and WHR were, respectively, 0.96 (95% CI: .94 –.97); .93 (95% CI: .91–.95); and .84 (95% CI: .78–0.89).

With the exception of the correlation between the assisted measurement WHR and diastolic blood pressure, all correlations were statistically significant (p < .05). Generally the assisted and self-measured correlation coefficients provided very similar estimates of the relationships’ with BMI, blood pressure, and blood glucose; and never significantly different from each other.

We sought to determine the extent to which a Malaysian community sample could provide accurate self measurements of waist and hip circumference and WHR, and the degree to which the self measurements could be used in studies of relationships with other health markers. The results of this study supported the use of waist and hip self measurement.

The measures had on average a small downward bias (around 1 cm), with 95% limits of agreement around ±6 cms around the downward biased estimate. The concordance for waist and hip measures was strong – 0.96 and 0.93 respectively. These results were similar or better than those reported by Lim and colleagues in their study of Thai students [15], and was broadly in keeping with other self measurement studies [22, 30, 31]. Also in keeping with those studies, WHR measurement was less accurate than WC or hip circumference measurement [29].

The greatest limitation of the study related to sampling. Like other community, occupational or university based studies where sampling was non-random, one is left to speculate about the generalisability of the findings [15, 21, 24, 31]. Specifically, what is the underlying population represented by the sample? Notwithstanding this limitation, the study does contribute additional data to an otherwise extremely limited evidence-base of non-European or North American studies [21–31]. Like self-reported BMI, with its known biases, self measurement of waist and hip circumference should not be mistaken for a gold-standard, and should not form the basis for a clinical assessment. However, for population-based studies where there are issues of either cost or personal privacy, self measurement is a credible, if potentially weaker alternative to assisted measurement.

The results suggest that self measurement of waist and hip circumference using constant tension measuring tapes provide a favourable alternative for population surveillance of central adiposity in a community setting. The relationships between self and assisted measures with other health markers were generally very similar.