Abstract
Abstract
Background:
Upper limb lymphedema is a possible consequence of the treatment for breast cancer. Accurate detection of swelling is important in implementing appropriate treatment. Currently used diagnostic cut-offs for excess volume have been chosen for ease of use and are not based on normative differences. The aim of this study, therefore, was to determine the normal inter-limb variance for healthy older women and identify statistically-based diagnostic cut-offs for both circumference and volume.
Methods and Results:
Two hundred and four healthy women, over the age of 40 years, with no history of treatment for breast cancer or lymphedema, underwent measurement of their upper limbs with a perometer. Using the associated software, the circumference of the limb was determined at a number of set points along the limb and the volume of the intervening segments recorded. Segment volumes were also calculated from the circumferential measurements using the formulae for a truncated cone and cylinder. The mean inter-limb difference found was small but a large range was seen for all of the circumference and volume measurements. Dominance was found to have a significant effect on the limb size. Regression analysis showed that an individual's age was negatively related to their inter-limb difference. Diagnostic cut-offs, set at three standard deviations above the mean, were determined.
Conclusions:
New circumference and volume criteria based on normative data, taking arm dominance into consideration, will allow for more accurate diagnosis of changes in limb volume, allowing treatment to be started and monitored appropriately.
Introduction
Clinical detection of lymphedema is typically based on measurements of arm circumference or measurements of volumes that are derived from the circumferences. Circumferential measurements, obtained with a tape measure, are clinically the most frequently used technique. 3 To determine the circumference, the tape measure is passed around the circumference of the limb at set intervals along the arm. The circumference measurement can be used either directly as a measure of limb size or converted into a volume measure using a geometric formula. These measurements of arm circumference and volume can also be obtained using a perometer. The perometer consists of a frame containing infra-red light emitting diodes, as well as light sensors located diametrically opposite, which can be moved along a fixed track. The patient's arm is placed in the middle of the frame and as the frame is moved along the track, the sensors detect the changes in light patterns, allowing an image to be created of the limb and its circumference to be measured at regular intervals from which the volume can be calculated. Perometry measurements have been shown to be easy to complete and to have good reliability.4,5
Measurements of volume derived from circumferences have good agreement with both the water displacement method of measuring limb volume, often considered the gold standard, as well as perometry. 6 Although volume measures have been derived using both geometric formulae for a truncated cone (frustum) and a cylinder, calculations based on the truncated cone were in slightly better agreement with the water displacement approach than the simple cylinder. 6 In addition, volumes derived from the perometer and the tape measure technique have high concordance. 5
Detection of differences in limb volumes with either circumference or volume measurements are based typically on the difference between the at-risk limb and the healthy limb. In the literature and clinical practice, a large number of diagnostic cut-offs have been used to determine whether swelling is present. Examples include: a 2 or 3 cm difference in one or more arm circumference measures;7,8 a total difference in circumferential measurements of 5 cm or 5%;9,10 an absolute inter-limb arm volume difference ranging from 125 to 200 mL, 7 or a 10%–20% inter-limb volume percentage change.7,8,11,12 The most widely accepted measure of arm lymphedema is volume 13 and the most commonly used volume cut-off is a 200 mL volume difference. 7
The origins of these diagnostic thresholds are uncertain.14,15 The mostly commonly referenced article for the original basis for the 200 mL criterion is that of Kissin et al., 16 although the authors do state that this difference was chosen arbitrarily for convenience. The origin of the 2 cm circumference difference, the most commonly used clinical diagnostic cut-off 7 , is equally vague. It would therefore appear that current diagnostic criteria for lymphedema are based on historical precedence or opinion, which may explain the lack of agreement among diagnostic criteria.10,17,18 This lack of evidence-based consensus contributes to the wide range of incidence rates reported in the literature for conditions giving rise to limb swelling such as lymphedema. 10
To date, the normative values for inter-arm differences have not been investigated using common clinical outcomes in healthy older women, the population at greatest risk for breast cancer-related lymphedema. The lack of evidence supporting the currently used criteria has been identified as being problematic, 19 leading to late diagnosis for some and false diagnosis for others.9,17 Both situations may lead to costly treatments that may otherwise been unnecessary and considerable emotional distress.17,20 Given the wide use of arm circumference and volume measurement for assessment and the lack of sound theoretical bases for currently used cut-offs, the aim of this study was to describe, in a group of healthy older women, what is the normal inter-limb variance and to determine statistically-based diagnostic cut-offs for both circumference and volume measurements. Given the high concordance of perometry-derived measurements with those using a tape measure, 5 both circumferential and volume measurements were obtained using a perometer.
Methods
Ethical approval
The study was approved by the Human Research Ethics Committee at each of the institutions where the study was conducted. Written informed consent was obtained from each participant prior to participation in the study.
Participants
Data were obtained from healthy participants who were recruited from a variety of community settings both in urban and regional areas in New South Wales, Australia. To be included, participants had to be over 40 years of age and have no self-reported history of breast cancer or lymphedema. In addition, participants were screened using a questionnaire for other conditions or recent upper limb injuries, as well as medication or supplement use that may have affected their fluid or limb volumes, such as beta blockers or diuretics.
Two hundred and four women participated in the study. It has previously been determined that 50–75 participants are necessary to determine normative data. 21 The sample size was determined to permit recruitment of sufficient people for the following 4 age groups: 40–49 years old, 50–59 years old, 60–69 years old, and 70+ years old. Participant characteristics are reported in Table 1. Fifty participants were recruited for all of age groupings except for the 70+ category. The participants' ages ranged from 40 to 86 years; the average age (standard deviation) was 58 (11) years. Body mass index (BMI) ranged from 15.8 to 48.4 kg/m2 with the 60–69 year age group showing the largest average BMI. Ninety-three percent of participants reported they were right hand dominant, which is similar to that observed in other studies. 5 Other participant factors such as ethnicity, or education level were not considered.
Mean (standard deviation) shown.
Protocol
Participants attended a single assessment occasion at which time their height and weight were measured and arm measurements obtained. Height was determined using a stadiometer, and weight using an analog scale, and body mass index (BMI) was calculated from these measures. Arm measurements were undertaken using perometery (Perometer, 1000M Juzo). The reproducibility of perometry has been established previously.4,5
Perometry measurements were undertaken as previously described. 5 The participant was seated with their arm in 90 degrees abduction, their elbow fully extended with their middle finger touching the end of the hand rest, palm facing down, and their thumb tucked into their hand. The hand rest was placed such that the participant had to reach slightly to maximize the length of arm measured. The participant's arm was centered over the perometer track and the frame was slid along slowly to the shoulder, returning again to the starting point. The other arm was then measured in the same manner. Measurements were completed a single time only. All measurements were completed by research assistants, with physiotherapy or exercise science backgrounds, who were trained and experienced in the correct positioning and use of the perometer.
Data processing
Circumferential data
Arm circumference measurements were determined manually using custom-modified Peroplus software™, commencing at the ulnar styloid. The participant's hand length was measured from the tip of the middle finger to the ulnar styloid, with this distance being used to determine the 0 cm starting point of the arm. Arm circumference measurements were then determined at this point and 10, 20, 30, and 40 cm proximally from the ulnar styloid for each arm. The 40 cm measurement was not available for 15 participants due to insufficient arm lengths as reported previously.45
Perometer calculated limb volumes
Limb volumes were determined using custom-modified Peroplus software™. The algorithm used to convert the circumferential measurements to volumes is unknown as it is proprietary information of the manufacturer. The software has a mean slice resolution of 4.7 mm but reports segmental volumes rounded to slices of either 4 or 5 mm nominal width. The volumes of these consecutive slices were resolved to 1 mm slices so that the volume of precise 10 cm segments could be determined. The volume of each 10 cm arm segment, starting from the ulnar styloid, was obtained by summating these consecutive 1 mm slice volumes. For the purposes of this study, the arm segments were designated as follows: the 0 to 10 cm was segment A, 10 to 20 cm was segment B, 20 to 30 cm was segment C, and 30 to 40 cm was segment D. Segment D was not available for 20 participants owing to insufficient limb length. For the participants with all four segments were available, the total limb volume was determined by summating the volume of all segments. Perometer calculated volume data were not available for 11 participants due to use of an older version of the software that lacked the ability to export the segmental volumes.
Derived limb volumes (based on arm circumference measures)
Two methods of geometric calculations were used to derive volume. The volume of each 10 cm segment, commencing at the wrist, was calculated based on the assumption that the segments approximated a truncated cone (frustum) and a simple cylinder.
The volume of the truncated cone was calculated using the following formula: 6
where V is the volume of the segment, C1 and C2 are the circumferences at each end of the segment, and h is the distance between the two circumferences, in this study, 10 cm.
The volume of a cylinder was calculated using the simple cylinder formula: 6
where V is the volume of the segment, Cavg (1,2) is the average of the circumference at each end of the segment, and h is the distance between the two circumference, 10 cm.
For both volume calculation methods, the volume was determined for each of the four segments and the total limb volume was calculated by the summation of the four segments for all except 15 participants who had insufficient arm length to provide the volumes of all four segments.
Data analysis
All data are presented as means±standard deviation for the dominant and nondominant arm separately for each of the volume calculations, circumferences, and limb segments. Volume data obtained by each method were compared using concordance correlation 22 and the limits of agreement method. 23 Paired t-tests were used to determine whether the circumferences or volumes of the dominant limb were significantly different to those of the nondominant limb.
The inter-limb differences were used to determine diagnostic cut-offs for each of the assessment methods. As limb dominance may impact on diagnostic cut-offs, the calculations were undertaken separately for both the dominant and nondominant limb. For calculations of the cut-offs for the dominant arm, measures for the nondominant arm were subtracted from those of the dominant arm for each participant, while for the nondominant arm cut-off, the reverse was undertaken. Mean and standard deviations of the difference for each variable was determined and cut-offs were established by using the previously reported approach based on the frequency distribution of the data24,25 in which the diagnostic cut-off is set at the mean plus three times the standard deviations, thereby encompassing >99% of the population.
Finally, regression analysis was undertaken to determine if age, side of dominance, weight, height, or BMI were related to the inter-limb differences in circumference or volume. Data were analyzed using SPSS for Windows (version 19, IBM) and Microsoft Excel (2007).
Results
The mean and standard deviations for the circumference measurements, the volumes of the 10 cm segments and whole limb for both the dominant and nondominant arms are shown in Table 2 for all participants.
n=193; 2n=185; 3n=170.
Mean (standard deviation) shown.
Comparisons of measurement methods
Volumes calculated by the perometer and truncated cone methods were well correlated with concordance correlation coefficients ranging from 0.866 to 0.987, depending upon which segment and limb (dominant or nondominant) were being compared. Similarly, the concordances between the perometer and the cylinder methods ranged between 0.851 and 0.988. The results for the truncated cone and cylinder calculated methods were highly similar with a concordance of 0.999 to 1.000. Despite all three methods being highly correlated, small but significant differences were present. For the total limb volumes, the mean bias between methods was 2.9% with the largest bias found between the perometer and cylinder for the dominant arm (5.0%). The bias was larger for the dominant arm than the nondominant arm. The limits of agreement (2 SD) between methods were larger when perometry was compared to the calculated volume methods than when the calculated volume methods were compared to each other. The largest limit of agreement for the whole limb measurements was −1.0% to 10.0% for the comparison of perometry with truncated cone for the dominant arm.
Effect of limb dominance
Paired t-tests between the dominant and nondominant circumference measurements revealed significant differences due to dominance for most but not all locations (Table 3). The mean differences in circumferential measurements between the limbs were small but significant for the 10 cm, 20 cm, and 40 cm measurements: 0.27 cm (95% CI=0.17–0.36 cm), 0.29 cm (95% CI=0.21–0.41 cm), and 0.17 cm (95% CI=0.03–0.31 cm), respectively. There was no significant difference between the limbs for the 30 cm circumference measurements.
p≤0.005; 2p≤0.03.
Paired t-tests for both the perometer and calculated volume measurements also revealed a number of significant differences between the dominant and nondominant limb. All of the perometer-determined segment volumes were significantly different between sides (p<0.03–0.001), while for the calculated volume measurements, only segment D was found not to have a significant difference between sides (segments A, B, C and whole limb p<0.04–0.001).
Diagnostic cut-offs
Cut-offs for diagnosis of swelling were based on three standard deviations above the mean difference between the limbs (Table 4). Diagnostic cut-offs were determined and differed depending on how far along the arm the circumference measure was taken and whether the dominant or nondominant arm was ‘at-risk’. For the dominant arm, starting at the ulnar styloid and moving proximally 10 cm, the cut-offs were 1.3 cm, 2.4 cm, 2.5 cm, 2.7 cm, and 3.1 cm, while for the nondominant arm they were 1.2 cm, 1.9 cm, 1.9 cm, 2.7 cm, and 2.7 cm.
Dominant arm mean differences are calculated by subtracting the nondominant arm values from the dominant arm, while the nondominant arm mean differences are calculated by subtracting the dominant arm values from the nondominant.
Three standard deviations above the mean difference between limbs.
Similarly, the volume based cut-offs for each segment increased from segment A through to D. Cut-offs for the perometer measurement were higher than those using the calculated volume method at 370 ml and 210 ml for the perometer and 290 ml and 240 ml for the truncated cone or cylinder conversion method for the dominant and nondominant arm, respectively.
Regression analysis found a small, but significant, relationship between the inter-limb volume difference for the whole limb and age irrespective of which volume method used (adjusted r 2 =0.039 and 0.065 for the perometer and truncated cone/cylinder, respectively). The standardized coefficient for age was −0.212 and −0.266 for the perometry and truncated cone/cylinder methods, respectively. Hand dominance, height, weight, and BMI were not related to inter-limb differences.
Owing to the significance of age as a variable in the regression, cut-offs based on age groups were determined (Fig. 1). While slight differences in the diagnostic thresholds were seen for the different age groups, a similar pattern of increasing cut-offs when moving proximally up the arm were for both the dominant and nondominant arms.
Diagnostic thresholds for the dominant
Discussion
To date, the normative size and inter-limb difference in healthy older women has been unknown, preventing appropriate detection of secondary upper limb lymphedema. Our present study is the first to quantify the normative inter-limb difference in older women. We found that, on average, the inter-limb circumference and volume differences were small, although a large range was present, particularly more proximally along the upper limb. We also found that for most but not all, there were significant differences between the dominant and nondominant limbs. In addition, we determined normative-based diagnostic cut-offs. The proposed new cut-offs are based on which limb was affected as well as describing differences in cut-offs dependent on the location of the swelling on the arm (Fig. 1A and 1B). We confirmed that the methods for volume measures (perometry versus calculated methods) were found to be related but not inter-changeable. 5 Finally, this study showed that the truncated cone and cylinder formulae could be used inter-changeably.
In the current study, we have proposed that an inter-limb difference three standard deviations from the mean be adopted for diagnosis of pathological volume changes. Thus, for women over 40 years of age, an inter-limb difference of greater than 380 mL is required for diagnosis of swelling of the dominant arm, when using the perometer. If the commonly used clinical cut-off of 200 mL was used, 8% of our sample would have been incorrectly diagnosed with pathological swelling, which is a similar proportion to what has been previously reported. 7 Only one other set of normatively-based volume cut-offs has previously been suggested determined using the water displacement method of measurement, 26 which is not commonly used in clinical practice. 3 Smoot and colleagues suggested a starkly different inter-limb volume difference cut-off of 75 mL based on area under the curve analysis 2 . However, in their cross-sectional study, the mean (standard deviation) inter-limb difference in their group of women without lymphedema was 8 (73) mL. Based on their study population's standard deviation, if their criterion of 75 mL was used, almost a third of the women without lymphedema in their study would be misdiagnosed as having lymphedema, while in our study, 53% would have been incorrectly diagnosed with lymphedema. This diagnosis could cause unnecessary psychological distress and negatively affect quality of life and lead to costly, unnecessary treatment. A cut-off that discriminates pathological differences from normal variance for the majority of the population is therefore necessary.
Limb dominance is not taken into account with the currently used cut-offs.7,27 Our results show that limb dominance should be a consideration, as for most regions of the arm there are significant differences between the dominant and nondominant limbs for both circumference and volume measurements. The effect of dominance has previously been found for other tools used in the diagnosis of lymphedema such as bioimpedance spectroscopy which has necessitated different cut-offs depending on which limb was at risk.24,25,28 It is therefore logical that circumference and volume cut-offs should also take into account limb dominance.
Age was found to have a small but significant effect on the inter-limb difference. This has been found previously with other tools such as bioimpedance spectroscopy. 29 Age-stratified thresholds may be necessary, but this should be confirmed by a study with larger sample sizes for each age group than were available in the present study, particularly for those 70 years and older. As the differences between the diagnostic thresholds for the whole group and the age groupings are not large and all continue to follow the trend of increasing cut-offs when moving proximally up the arm, the whole group thresholds would be appropriate for use in women over the age of 40 years old.
This study does have a few limitations. Rather than measure the limbs with a tape measure at 10 cm segments, we chose to use the perometer rather than the more commonly used tool of the tape measure. The basis for this decision was for efficiency as we recruited our participants at meetings and public gatherings. However, we do not think that this affected our results, as the perometer has previously shown to be well correlated with measurements derived using a tape measure.4,5 It has been suggested that the perometer and the tape measure could be used interchangeably if a small correction factor were used to convert the tape measure measurements to those of the perometer. 4 This may change the cut-offs slightly. Another limitation was that hand dominance was determined by asking the participant which arm they felt was dominant, using the clarification of which hand they wrote with if needed. This may not be the best way to determine which hand is actually dominant, as it has been suggested that the Edinburgh Handedness Inventory is a more valid method to determine handedness. 26 Clinically, however, a patient's limb dominance is usually self-determined, as was done for this study.
While this study provides strong evidence for diagnostic cut-offs in healthy older women, further study is needed. For this research, a three standard deviation above the mean cut-off was chosen for the diagnosis of swelling, as it has been previously deemed appropriate elsewhere for the diagnosis of lymphedematous swelling. 24 It has been suggested, though, that this may be too conservative a criterion to use. 25 While this criteria has been accepted and supported for other tools used in the diagnosis of swelling, 25 others have suggested that two standard deviation above the mean may be more appropriate. 26 If a mean plus two standard deviation cut-off were to be adopted, our data indicate that for volume these would be 275 mL and 115 mL for the dominant and nondominant arm, respectively, when measured by perometry. The respective values for truncated cone measurements of volume would be 200 mL and 150 mL. For this sample, these lower cut-offs would have led to 2% being incorrectly diagnosed as having pathological swelling. In addition, the group of interest for this study was healthy women over the age of 40. It is unknown whether these results and cut-offs would be appropriate for younger women or for men of any age. It has previously been found that the majority of tasks completed by older adults used both arms, 30 which may result in smaller inter-limb differences than other populations who may complete more unilateral tasks leading to a larger inter-limb difference. This was found in this study, with the regression analysis showing that with increased age, the inter-limb difference decreased. Further study is needed to determine the utility of these measures across these groups.
Conclusion
Correct diagnosis of secondary lymphedema is very important in avoiding unnecessary treatment. Current criteria used in lymphedema have not been based on normative differences but rather on convenience. The wide range of inter-limb differences highlights the importance of obtaining pre-operative measurements where possible. However, when this does not occur, new circumference and volume criteria based on normative data, taking arm dominance into consideration, will allow for more accurate diagnosis of changes in limb volume, allowing treatment to be started and monitored appropriately.
Footnotes
Acknowledgments
Thank you to Sandra Soul and the other research assistants who assisted with data collection. Thank you to Mungovan Breckenridge Physiotherapy and Associates and Zonta International for their assistance with this research.
Author contributions: E.S. Dylke contributed to the study conception and design, data collection, analysis and interpretation of data, and manuscript preparation; J. Yee contributed to data collection and manuscript preparation; L.C. Ward contributed to the study conception and design, analysis and interpretation of data, and manuscript preparation; N. Foroughi contributed to data collection and manuscript preparation; S.L. Kilbreath contributed to the study conception and design, analysis and interpretation of data, and manuscript preparation.
Author Disclosure Statement
ESD was supported by a biomedical postgraduate scholarship from National Health and Medical Research Council (Australia). SLK was supported by a career fellowship from the National Breast Cancer Foundation (Australia). The authors have no potential conflicts of interest.