Introduction Body mass index (BMI) is the pragmatic measure to assess children's obesity clinically and BMI charts are widely used for counselling families about children's weight management over time.
Aims To explore the variability in clinicians' interpretation of BMI patterns and to ascertain the diagnostic accuracy of their judgement by relating it to change in body composition by dual-emission x-ray absorptiometry (DXA).
Methods Data from 70 children who participated in a trial of a weight management programme for obese children were analysed. BMI was plotted on UK 1990 charts at baseline, 6 months and 12 months, and four clinicians experienced in obesity management independently scored the charts on a five-point scale for how successful children were in tackling their obesity over a 6-month period. Scores were compared with change in BMI, fat mass and lean mass z-scores as measured by DXA.
Results 54 children (aged 8–15 years; BMI z-score 2.93 (SD 0.48)) had simultaneous BMI and DXA scans performed, giving 104 pairs of measurements 6 months apart. There was good consistency between clinicians' scores for weight management and these related well to change in BMI and fat mass z-scores, but not lean mass z-score. They reported that measurement proximity to centile lines and crossing of lines influenced their confidence in making a decision and change in severe obesity was harder to judge as higher centile lines are so far apart.
Conclusions BMI charts are useful for assessing children's attempts at weight management, and provide a reasonably accurate indication of change in body fat. Recommendations are made regarding BMI chart design and guidance in interpreting measurements.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
With the knowledge that childhood obesity is far from benign there has been an increasing emphasis on the importance of weight management.1 Although it is clear that body mass index (BMI) is not a precise measure of obesity, in practice it is the only practical measure to use in clinical settings for this purpose.2
In 1995, BMI charts were introduced based on the UK 1990 growth references3 to provide clinicians with a tool to determine how overweight (or underweight) a child was. The highest centile was the 99.6th, but it rapidly became clear that this was inadequate given the number of children with extreme obesity. ‘Management’ charts4 were then produced (see figure 1) providing clinicians with centile lines equivalent to 3.5 and 4 SD scores. These are now in general use in obesity services across the UK and a working party has been convened by the Royal College of Paediatrics and Child Health (RCPCH) to see how they might be improved.
What is already known on this topic
▶ For pragmatic reasons, body mass index (BMI) is the accepted measure for obesity in the clinical setting.
▶ BMI charts based on cross-sectional data have been available for use in children since 1995.
▶ There has been no evaluation of how clinicians interpret BMI patterns over time or how their interpretation relates to true change in adiposity.
What this study adds
▶ There is good agreement between experienced clinicians when interpreting BMI charts for success and failure in weight management.
▶ Clinicians' judgement correlates reasonably well with change in obesity as measured by change in BMI and fat mass SD scores.
▶ The findings from this study allow recommendations to be made about BMI chart design, and guidance in interpreting BMI change in obese children.
The UK 1990 charts were compiled from cross-sectional data. They were subsequently validated for monitoring height longitudinally.5 However, it is far from evident that the BMI charts are adequate for tracking change in BMI. BMI charts are based on cross-sectional data, but the numbers of children who contributed to the extremes of BMI (>3 SD) were small.
There has been extensive research comparing the diagnostic accuracy of BMI with gold standards of body fat as measured by dual-emission x-ray absorptiometry (DXA) or MRI scanning.1 2 A recent paper6 has also determined the reduction in BMI z score required to bring about change in morbidity markers such as blood pressure, lipids and insulin levels. However, despite the widespread use of BMI charts, paediatricians' clinical judgement has not been evaluated in terms of their interpreting children's growth patterns. This is important as clinicians, when seeing a child for obesity over time, base their guidance and encouragement on the change in BMI when plotted on a chart.
Our aim was to explore the variability in the interpretation of BMI patterns by clinicians who are experienced in managing paediatric obesity and to ascertain the diagnostic accuracy of their judgement by relating it to change in fat mass by DXA. In so doing we hoped to develop some practical guidance for clinicians when interpreting BMI charts and to provide information that would be of value to the RCPCH working group in redesigning the BMI charts.
The anthropometric and body composition data from 70 children who had participated in a randomised controlled trial of WATCH IT, a weight management programme for obese children,7 were analysed. Measurements had been taken by a trained researcher in the paediatric outpatients department of the Leeds General Infirmary at baseline and after 6 months and 12 months of participation in the trial. Height was measured to 0.1 cm using a wall-mounted Seca stadiometer (Vogel and Halke, Hamburg, Germany), the two measures being averaged. If they differed by >0.5 cm a third measure was taken, and the average of the closest two used. Weight was measured in light clothing with no shoes (to 0.1 kg) using a calibrated Seca digital weighing scale (Vogel and Halke). Total body fat and percentage of body fat were measured by DXA (Lunar Prodigy; GE Medical Systems, Madison, Wisconsin, USA) on the same occasion.
For the purpose of this study, children who had received the intervention were pooled with those who were waiting list controls. Their BMI measurements were plotted on the UK 1990 growth charts. Four clinicians with an interest in obesity (a paediatric endocrinologist, a community paediatrician, a paediatric specialist registrar and a specialist endocrine nurse) were asked to individually assess each child's growth pattern, and categorise it on a five-point scale. They were instructed to look at the BMI change from baseline to 6 months and from 6 to 12 months, and to indicate how much each child's obesity had changed over each 6-month period, guided by their thoughts as to how much they might congratulate or be disappointed at the family's efforts at weight management. The five categories were as follows: marked increase in obesity; some increase in obesity; no real change in obesity; some decrease in obesity; marked decrease in obesity.
Examples of three children's BMI charts are shown in figure 2, with the scores they received for success in weight management allocated by the clinicians.
To allow comparison of measurements of children of different age and gender, BMI z-scores3 were calculated using the Excel add-in LMSgrowth8 (z-scores indicate the relation of a measurement to those of children of the same age and gender, with ‘0’ corresponding to the 50th centile, and ‘+1’ or ‘−1’ representing 1 SD above or below the median, respectively). Fat mass and lean mass z-scores were calculated similarly using body fat reference standards produced by Truscott J and Crabtree N. Unpublished fat mass index and fat free mass index DEXA data. Leeds General Infirmary, Leeds, UK (personal communication 2008).
The four clinicians' scores were plotted against change in BMI z-score, change in fat mass z-score and change in lean mass z-score to allow a visual presentation of their agreement in judging children's successful or unsuccessful attempts at weight management. The analysis deliberately ignored the repeated measures design in individuals, and the allocation of individuals to either the control or intervention arm of the trial because the aim was to compare BMI change with body fat change over distinct 6-month periods, irrespective of changes at other times.
To gain an understanding about the way clinicians came to their judgement, BMI trajectories and scores were then viewed qualitatively by the senior clinician (MCJR) alongside reflections written down by the clinicians on completing the exercise.
Ethics approval was obtained from the Leeds (West) Research Ethics Committee. Written parental consent and child assent was obtained for participation in the trial of WATCH IT and subsequent analysis of the data.
Fifty-four children aged 8–15 years had simultaneous BMI and DXA scans performed on two or three occasions, each 6 months apart, with a total of 104 pairs of measurements. They were clinically obese at baseline with mean BMI z-scores of 2.93 (SD 0.48), well above the 99.6th centile. Baseline characteristics and change over time are shown in table 1.
Figure 3 shows the four clinicians' assessments of the children's BMI patterns plotted against change in BMI z-score over time. The scores are clearly grouped. There was clear differentiation between scores ≤2 (some or marked decrease in obesity) and ≥4 (some or marked increase in obesity) about the line of zero BMI z-score change. The overall agreement was good with full agreement for 42 scores, 58 differing by one point and only 4 by two points.
Figure 4 shows the clinicians' scores in relation to the children's change in body fat as indicated by the change in fat z-score. Once again there was good agreement among the clinicians as shown by the line up of the symbols. However, in comparison to figure 2, the grouping was less tight, indicating that their judgement correlated less well with true adiposity.
By contrast, figure 5, which illustrates the correlation of scores with change in lean mass, shows no association. This is important because it indicates that clinicians' scoring did not relate to changes in lean mass. Also note that the range of change in z-score is about 50% greater than that seen in BMI or body fat.
Table 2 provides information on the clinicians' views on completing the exercise of scoring children's BMI charts. Crossing of centile lines clearly influenced their decision as to whether or not a child had exhibited a marked change in obesity (up or down) and they tended to give a more extreme score when centiles were crossed. They found that scoring severely obese children was more challenging, in part due to the paucity of centile lines on the upper part of the chart. Children on the lower centiles needed smaller changes to be considered successful. A reduction in raw BMI was not required to consider a child had achieved a reduction in obesity, and no change in BMI over 6 months was seen by the clinicians as being an indication of success.
BMI charts have been introduced into paediatric clinics and weight management programmes to help clinicians assess a child's obesity and measure change over time. They have followed the format of height and weight charts, but unlike height and weight charts they have not been validated for clinical use, nor guidance provided about how to interpret change over time.
In our study four experienced clinicians studied the change in BMI over 6 months for 54 children, and indicated how they would communicate the change to parent and child. Our results show remarkable consistency among the clinicians, with little discrepancy in the scores they gave. Their judgement also correlated well with change in BMI z-score.
Not surprisingly there was somewhat less correlation with body fat as measured by DXA, which is a superior indicator of obesity.1 2 Nonetheless the clinicians' score was a reasonable predictor of change in fat mass, though not perfect, so some caution is needed when counselling patients about the success or failure of attempts at weight reduction.
An important finding is that clinician scores did not relate to changes in lean body mass, with the spread being greater than for fat mass. The reasons for this are not entirely clear, but may relate to differences in stage of puberty. This means that the BMI chart provides no guidance as to the likely change in lean mass over time, so it is not helpful for judging the impact of, for example, a claimed increase in physical activity.
The clinicians' reflections on scoring children's BMI charts revealed some important insights. They used the centile lines for coming to conclusions, and crossing centile lines clearly influenced their judgement. They reported that judging change in severely obese children was more difficult because the higher centiles are much further apart, which also resulted in a tendency to consider smaller changes in the lower centiles as more substantial. This would have a clear influence on how encouraging, discouraging or even disbelieving the clinician would be about weight management attempts. It was interesting to see that an actual reduction in raw BMI was not required to consider a child had succeeded in reducing obesity. Indeed, if no change in BMI occurred over 6 months, clinicians regarded that as being a marker of success.
In considering the results of our study it is important to reflect on the sample of children who contributed data. There was a wide range of obesity with all children being appreciably obese; BMI z-scores ranged from 2.4 to 3.8, with only 12 below the 99.6th centile at the start of the study. Only seven achieved a 0.25 reduction in BMI z-score, a change that has been associated with improved morbidity markers in children.6 The children were in a range likely to be seen in paediatric obesity clinics. Their success overall in weight reduction was poor relative to changes seen in randomised controlled trials of weight reduction programmes,9 although it has to be remembered that children in the control arm were included too. The level of weight change observed is in keeping with observations in paediatric outpatient clinics,7 although it is noteworthy that one paediatric centre has reported greater success.6
This study allows us to make some recommendations about the format of BMI charts. It seems that clinicians rely heavily on centile lines to interpret the charts, so that additional centile lines would be helpful above the 99.6th centile (perhaps at +3 SD and +3.75 SD). It would also be helpful if the time scale was expanded in line with height and weight charts so that BMI points are not plotted too closely together in the usual time interval allowed for repeat measures. It has to be noted too that two children's measurements could not be plotted as their BMI exceeded 49 kg/m2, off the scale of the chart. This is unfortunately not uncommon, and as the charts are intended for the extremely obese the upper limit needs to be extended, possibly to 60 kg/m2. Lastly, the current charts do not provide information about the use of BMI charts and their interpretation. Based on this study and clinical experience we propose some guidance, as shown in box 1, that we hope will be helpful.
Box 1 Guidance for clinicians in interpreting body mass index (BMI) charts
▶ BMI charts are a good tool for demonstrating how far outside the healthy BMI range a child is.
▶ BMI should not be assessed more frequently than every 4 months because the measurement error for height increment (which is used in its calculation) is too big over shorter periods.
▶ BMI is not an ideal measure of adiposity and interpretation of BMI change (or lack of it) needs caution. Change in BMI is only an approximation of the real measure of interest which is change in body fat.
▶ Caution is needed before assuming that change in BMI is due to an increase in lean mass.
▶ Crossing centile lines is a reasonable indication that there has been change in body fat.
▶ Tracking along an imaginary inter-centile line is a reasonable indication that percentage of body fat has remained constant.
▶ The more obese the child is, the harder it becomes to make conclusions about the interpretation of BMI patterns.
▶ Changes appear more dramatic at lower centiles than at higher centiles.
▶ One should be aware that actual crossing of centile lines tends to confer greater confidence in interpretation.
▶ It is important to set realistic goals. A reduction of 5–10% in BMI over time confers health benefits. (Depending on the severity of obesity, this equates to approximately one-third to two-thirds of the distance between two centile lines.) However, holding BMI steady or reducing its rate of rise is a great achievement for most young people.
In summary, our results show that experienced clinicians are consistent in their reading of BMI charts. Their judgement about success or failure at weight management correlates well with BMI change and is reasonably predictive of change in fat mass; it does not relate to change in lean mass. Clinicians found it harder to interpret BMI change in severely obese children and would appreciate extra centile lines above the 99.6th centile to guide their judgement.
Dr Ianthe Abbey, Dr Talat Mushtaq and Jenny Walker for interpreting 54 BMI charts, and all the children who contributed data through their participation in the WATCH IT trial.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.