The development of new monitoring technology has assisted in allowing high-quality data to be recorded in a variety of free living active situations (Achten and Jeukendrup, 2003; Jobson et al., 2009). New measuring technology can collate information on multiple integrated physiological and activity variables which can be assessed in real-time or downloaded post-performance. Reproducibility, or repeatability, of data using new measuring technology is crucial if advancement of ecologically valid assessment of activity is to continue. The Bioharness^™ (Version 1), can collate information on multiple integrated physiological and activity variables which can be assessed in real-time or post-performance. The Bioharness^™ simultaneously measures five variables (i.e. heart rate, breathing frequency, skin temperature, activity and posture), which can be monitored wirelessly in real time or downloaded from the device after the activity. Previous literature supports the use of each individual variable which is integrated in to the device; Heart rate (HR) through chest mounted electrodes (Grossman et al., 2006; Kent et al., 2009; Leger and Thivierge, 1988; Macfarlane et al., 1989), Breathing Frequency (BF) through respiratory inductive plethysmograghy (Grossman, et al., 2006; 2010; Kent et al., 2009; McCool et al., 2002; Witt et al., 2006), infra-red Skin Temperature (ST) (Burnham et al., 2006; Gant et al., 2006; Hershler et al., 1992; Matsukawa et al., 2000), Tri axial Accelerometry (ACC) (Brage et al., 2005; Powell and Rowlands, 2004; Rowlands et al., 2003) and Posture (P) (inclinometry) (Hansson et al., 2001; 2006) both the latter variables using piezoelectric technology. The Bioharness^™ device is being used within a variety of applied free living situations including the assessment of physical activity and within the emergency services for both rescuers and those being rescued. There is a lack of peer reviewed literature considering the reproducibility of the Bioharness^™ device with only the breathing frequency variable tested (Hailstone and Kilding, 2011). Measurements made by new technology in any environment must have known clarity as to what variability may exist (Atkinson and Nevill, 1998; Welk et al., 2004) and to our knowledge there has been no peer reviewed paper published detailing the reliability of all five variables associated with the device. Therefore, the aim of this paper was to assess the reliability of each variable within the Bioharness^™ device in an exercise based laboratory situation.

To assess the reproducibility of the Bioharness^™ variables appropriate assessment protocols were identified. A between (n = 10, using 1 Bioharness^™ device) and within subject (n = 1, testing 4 different Bioharness^™ devices) design, using a repeated treadmill protocol, allowed the assessment of ST, HR, BF and ACC, with the latter 3 variables being assessed at different velocities. P variable was assessed as a whole data set through a separate mechanical protocol. All data collection was synchronized to one timeline linked to a laptop computer. For consistency, a standardised technique for the fitting of all equipment was completed by one experienced researcher throughout the duration of the study.

The Bioharness^™ (Version1) is worn against the skin by the participant via an elasticated strap attached around the chest (50 g, 50 mm width). The monitoring device (weight 35 g, 80x40x15mm), which attaches to the front of the chest strap, acts as a data logger or transmitter, has a memory of up to 480 hours and battery life of up to 10 hours in logging mode. Five variables are measured simultaneously, time stamped and exportable to Excel. HR data is captured through electrode sensors housed within the chest strap (i.e. detecting R wave forms) at 250 Hz and reported as beats per minute (b·min^-1). BF is provided using a capacitive pressure sensor (18 Hz) that detects circumference expansion and contraction of the torso producing an output as breaths per minute (br·min^-1). Tri axial ACC, using piezoelectric technology (i.e. cantilever beam set up) samples at 18 Hz and reports 1Hz in counts per second (ct·sec^-1). It is a micro electro-mechanical sensor accelerometer with a capacitive measurement scheme and is sensitive along 3 orthogonal axes (vertical (x), sagittal (z) and lateral (y)). Acceleration data is monitored in gravitational force (g) in a range of -3 to +3 g on each single axis or as Vector Magnitude Units (VMU) which is an integrated value over the previous 1 second epoch:

The P variable uses similar piezoelectric technology as described. Acting as an inclinometer, data in angular degrees (°) ranges between -90° and +90°, it assesses the degrees the monitoring device is "off vertical". ST data is collected through an infrared sensor behind a clear window on the apex of the monitoring device. It records peripheral skin temperature at the inferior sternum. This sensor reports data (1 Hz) in degrees Celsius (°C).

After securing local institutional ethical agreement 10 male volunteers (age 20.5 ± 2.1yrs, body mass 70.4 ± 9.4 kg, body stature 1.77 ± 0.10 m) who were physically active, injury free and familiar with using a treadmill consented to participate. Participants were asked to refrain from consuming alcohol, caffeine, keep hydrated and rested 24 hours before testing. On arrival to the laboratory anthropometrical data were taken with stature (Seca 214, Birmingham, UK) and body mass (Seca 761, Birmingham, UK) measured (Stewart and Eston, 2007).

Reproducibility of these variables were assessed by participants completing an adapted discontinuous incremental treadmill protocol (Rowlands et al., 2004). In a thermo-neutral laboratory environment (23.9 ± 1.7°C) the protocol consisted of 6 discontinuous incremental stages: rest (0km·h^-1), walking (4 and 6 km·h^-1); and running (8, 10 and 12 km·h^-1) performed on an electronically driven treadmill (HP Cosmos Mercury, Germany). Stages lasted a total of 8 minutes; 2 minutes rest, 4 minutes being active (i.e. walking or running) followed by 2 minutes recovery. Data was collected every 5 seconds for the last 90 seconds of each of the respective active stages. Participants were fitted with the Bioharness^™ 15 minutes prior to test commencing and remained on the treadmill throughout. The retest was completed 5 days from the date of first test, at the same time of day as test 1, with participants adhering to the same 24 hour pre-test protocol described previously.

In a controlled procedure, reproducibility of P data was tested by securing Bioharness^™ devices to an inversion (i. e. tilt) table (F500III, STL International) was moved through 160° as noted elsewhere (Bernmark and Wiktorin, 2002). The tilt table was calibrated (to 0°) using a spirit level and then moved through a 160° (+80 to - 80) at 10° intervals, pausing for 10 seconds, at each interval allowing data to be recorded. This process was then repeated.

Data was exported to statistical software packages (Excel Microsoft Windows, USA; SPSS v17, SPSS Inc, Chicago, USA) for analysis. When assessing reproducibility a range of statistical procedures have been cited in combination with descriptive data are available for researchers providing a comprehensive summary (Atkinson and Nevill, 1998; Hopkins et al., 2009; Kent et al., 2009; Nunan et al., 2008; Sandercock et al., 2005). Reliability of the data was assessed through the use of descriptive statistics (mean ± standard deviation (S)), Change in mean, 95% Confidence Limits and reliability statistics, Coefficient of Variation, Inter Class Correlations. The change in mean and associated 95% confidence limits will provide an indication of absolute variation between the data sets. Coefficient of variation (CV) expresses the S as a proportion of the mean, is considered a dimensionless statistic and therefore easier to compare variation between protocols. A somewhat arbitrary CV acceptable boundary of < 10% for reliability has been cited in some papers though this is not accepted unanimously in the literature (Atkinson and Nevill, 1998; Currell and Jeukendrup, 2008; Hopkins, 2000). Information on the relationship between sets of data is provided by correlation coefficients. Intra class coefficients are more sensitive to systematic bias and also can be used for multiple retests so have been preferred within reliability studies (Atkinson and Nevill, 1998; Hopkins, 2000). Boundaries for correlation statistics are not confirmed though amalgamated thoughts of Leger and Thivierge, 1988 and Hopkins, 2000 suggest; r > 0.9 Excellent or very strong, r = 0.7 - 0.9 Very Large, r = 0.7 - 0.5 Good to moderate, r < 0.5 Moderate to minor. All of these statistics analysed collectively will provide a clear overview on the reproducibility of data.

Characteristics of the data set were considered and appropriate statistical procedures followed thereafter. After plotting the between subject predicted against the residuals for HR and BF (Figure 1), data were considered to be non-uniform (i.e. heteroscedastic or not normally distributed) so were logarithmically (log) transformed in order to provide a true interpretation (Atkinson and Nevill, 1998; Hopkins et al., 2000; 2009). Descriptive data for these variables were reported in absolute values and reliability statistics presented log transformed which was determined in order for comparison with other studies to occur, the majority of which report absolute data.

Previously research assessing reliability of a monitoring device has removed data sets when data is clearly erroneous in the belief that a technical breakdown has occurred with the system (Leger and Thivierge, 1988). Analysis completed which includes highly erroneous data sets would possible reduce the practical usefulness of the results especially if this data was linked to a small clearly identifiable number of participants. The reporting of data removal (i.e. cleaning) has been used as additional evidence for reproducibility with high volumes of data being removed possibly reducing the reliability of the device. Therefore reporting of raw and clean data sets was completed on HR and BF data where some highly erroneous data was noted. Based on estimated maximal values of each physiological variable (McArdle et al., 2009), day-to-day biological variation (Achten and Jeukendrup, 2003) and considering other research (Hailstone and Kilding, 2011; Leger and Thivierge, 1988) the following data set removal criteria was established; If absolute mean of a data set presented a difference of ± 20 b·min^-1 for HR, ± 7 br·min^-1 for BF in comparison to equivalent data from the specific stage, the data was removed.

Between subject results (Table 1) for the whole data set note, low coefficient of variation (CV) and very strong relationships for P and ACC (p < 0.01). Less reliable variables are, ST, HR and BF, with ST variable having low coefficient of variation and moderate relationships. HR and BF present moderate relationships and a large CV.

HR results (Table 2) noted strong to very strong relationships (r > 0.84, p < 0.01), a lowering CV (< 6.2) and change in mean < 3.16 b·min^-1 with velocity ≤ 8 km·h^-1. Reproducibility of data decreased at velocities at ≥ 10 km·h^-1 with increases in change of mean (> 14.01 b·min^-1), CV > 24.7 and limited relationships in data.

BF data (Table 3) presents a weak relationship (r < 0.51), with elevated CV values (ranging 16.8 - 21.9). The change in mean remains < 1 br·min^-1 and this value reduces from rest to the active stages.

Data considered to be highly erroneous was removed following the procedure described earlier. HR (n=6) and BF data (n = 8) produced data (Table 4) mirroring trends from statistics seen at lower velocities. Considering this clean data set with the other velocities, HR change in mean remained < 3.16 br·min^-1, CV < 6 and strong to very strong relationship (r > 0. 84) were noted. BF data continued with < 1 br·min^-1 for change in mean, CV presented its lowest values at 10 and 12 km^-1 and a low to moderate relationship were identified.

At rest, ACC data (Table 5) presented the least reliable data with largest change in mean, largest CV and weakest relationship. As velocity increased, the change in mean reduced and became consistent, CV decreased (< 9.3) and moderate to strong (r > 0.66) relationship are reported.

General findings for intra (Table 6) and inter (Table 7) device reliability presented mainly strong statistics for HR, ACC, P and ST (r > 0.89, p < 0.01; CV ≤ 10.1). BF variable performed less effectively in comparison (r < 0.72; CV 11.4 - 17.4). No data was considered highly erroneous so analysis includes all data.

Further velocity specific intra and inter device results note low CV for HR (< 7.3) and ACC (< 10) with a general trend of decreasing variance with increasing treadmill velocity. At 0 km·h^-1, ACC presented high inter and intra device variance (CV range ~50 - 130) which reduced at the onset of activity. BF presented CV values < 11.4 with one exception during the inter device analysis (CV = 17.8). Correlation values for all variables were predominately low (r < 0.70; p < 0.05) with exception of one HR result (10 km·h^-1, r = 0.86).

Multivariable monitoring devices within sport and exercise can now provide time synchronised data which possibly could allow for further insights in to performance. Ensuring that a comprehensive precision of measurement assessment has occurred will allow for an understanding of the variability which exists and is a crucial step in achieving credibility in the market place (Welk, et al., 2004). The aims of the study were to assess the reliability of the Bioharness^™ monitoring device due to limited information being available on this issue.

Overall results suggest that, the Bioharness^™ produces adequately reliable data for HR, ST, ACC and P, with the latter two variables presenting the most accurate data. Erroneous data at higher velocities for HR and BF variables sets suggests caution should be applied to data collected during activities involving movement above 10 km·h^-1. For both BF and HR variable, between subject reliability improved after data cleaning at higher velocities. ST achieved the least test-retest reliability between subject, though produced stronger results in the within subject design.

HR data suggests adequate reproducibility across both testing designs at moderate velocity (≥ 8 km·h^-1) (Table 3). Considering the between-subject design, before data cleaning, there is weaker reproducibility in the data at higher velocities (≥ 10 km·h^-1) which is linked to an increase in highly erroneous data being produced. This latter phenomena was not apparent in the within subject data with reliability statistics being strong (r > 0.99; CV < 6.8) throughout all velocities. After erroneous data had been removed at ≥ 10km·h^-1, the between subject data presents equivalently strong results with slight trend of decreasing CV with increases in velocity, as previously reported elsewhere (Achten and Jeukendrup, 2003). In an arguably less intense yoga environment the results improve on correlations (r= ~.6) and match CV data (1.9 - 5.7) found for the LifeshirtTM, a multi variable assessment device (Grossman et al., 2006; Kent et al., 2009) and also is similar to unpublished data (CV 1.7 - 6.7) from our laboratory on the Polar HR monitor (T31, Polar Electro, Kempele, Finland). There is an expectation that this data should be within credible limits as monitoring HR telemetrically through electrodes housed within a chest strap has had over 20 years of development (Achten and Jeukendrup, 2003; Laukkanen and Virtanen, 1998).

Strong reproducibility data for ACC variable was noted in all testing scenarios and supports the notion that piezoelectric technology within the device can be deemed reliable (Table 5, 6">6 and 7">7). A low change in mean, low CV (< 8) and very strong relationships (r > 0.99) match or exceed previous research suggesting the ACC provides reliable data within the testing environment (Brage et al., 2003; Powell et al., 2003; G. J. Welk et al., 2000). ACC data at rest (i.e. 0 km·h^-1) was not included in the overall analysis as during pilot testing this data was inconsistent. During the rest stage inevitable slight irregular motion of the subject was registered as an activity count. This erratic non-rhythmical data production led to spurious variance in comparison to the remainder of the ACC data set. Piezoelectric elements are more effective in dynamic rather than a static mode (Chen and Bassett, 2005) and the data notes a lowering of CV as treadmill velocity increases (4 - 12 km·h^-1) which also corresponds to findings for other reliable ACC such as the RT3 (Powell, et al., 2003) and ActiheartTM device (Brage et al., 2005). Additional evidence from a study incorporating a free movement trial (e.g. sit-to-stand task) produced a wide range CV (8.7 - 25.6) between subjects which further corroborate this technical finding within the Bioharness^™ (Brage et al., 2005; Powell and Rowlands, 2004).

P variable, as assessed by a tilt table, produced good reliability statistics (r > 0.99; CV < ~10). Additional analysis was completed during the treadmill activity comparing posture during the within subject protocol which produced additional evidence that this variable is reliable (Table 1, 6">6 and 7">7). There are other tools to measure angular degrees in humans and even though it has been reported digital inclinometers are more reliable than goniometers they are not extensively used due to the expense (Venturni et al., 2006). Other research on this variable using similar technology has also demonstrated good precision of measurement (Hansson, et al., 2006). The same piezoelectric technology is used within P and ACC variable and both have demonstrated good reproducibility data.

Across both experimental designs the wider statistical analysis suggests the BF variable produced less reliable data in comparison to the other variables (Table 1, 3">3, 4">4, 6">6 and 7">7). As with the HR variable, higher running velocities (≥ 10 km·h^-1) lead to an increase in erroneous data occurring. After data cleaning, variance during the active stages (> 4 km·h^-1) seems to remain constant with slight decreases in CV at the higher velocities. Weak relationships were identified and CV values ranged from 21.9 at rest to a low of 10.4 at 10 km·h^-1. Interestingly, in comparison Hailstone and Kilding, 2011 noted higher absolute test-retest differences (< 2.8 br·min^-1) but found stronger relationships in data (r = 0.86 - 0.96) when they assessed reliability BF of the Bioharness^™. A difference in methodology and analysis may explain some aspects of these results, also they fail to identify the Bioharness^™ version used. Later versions of this technology may well use improved or different data processing algorithms. Comparing the data to other corresponding respiratory inductive plethysmography technology, a non-active environment presented stronger test-retest relationships (r = ~ 0.8) (Grossman et al., 2006) and a repeated within subject treadmill test for the LifeshirtTM device reported CV ~10 (Kent et al. , 2009). Weaker data from this study of the Bioharness^™ could be linked to the respiratory inductive plethysmography technical set up of the device. The LifeshirtTM adopts a 2 degree (i.e. 2 measuring band) model allowing thoracic and abdominal movements to be considered in producing respiratory data (McCool et al., 2002) in comparison to the one thoracic measuring band used within the Bioharness^™. The BF CV results suggest quite high variance especially when considering respiratory values during calibration should be within ± 3% (Zeballos et al., 2003) and evidence has presented lower CV (9.1) within maximal testing (Garrard and Emmons, 1986). Though Kent et al., 2009 reported high CV (~17) for breath-by-breath data gained from a Cosmed metalyser which also corresponds to unpublished CV data from our laboratory using a Cortex 3B metalyser (Cortex Medical, Germany). It seems that BF may be a physiological variable with higher variance, especially if analysed breath-by-breath and discipline specific data processing methods with regards to data averaging are seemingly not standardised, so could influence outcomes presented elsewhere (Kent et al., 2009).

Repeatability of ST in a thermo-neutral environment during the treadmill activity produced somewhat equivocal results (Table 1, 6">6 and 7">7). ST noted lower relationships in data for between subject design (r = 0.61) than reported in other research (Burnham, et al., 2006; Gant, et al., 2006) though a low CV (3.7) was maintained. For the within subject design, except for one device, relationships were strong (r > 0.89) which coupled with a low CV (< 1. 7) suggest the device attains good reliability. The difference in reliability between the two testing designs could be related to the positioning of the infra-red device relative to the subject. In the single subject (intra device) design, when the subject somatotype was standardised, data were more consistent. Previously the infra-red device placement, including lens angle and distance from skin, have been identified as important in attaining credible data and could have influenced the inter subject data collection (Hershler et al., 1992; Matsukawa et al., 2000).

It is important to identify if technical breakdown of equipment occurs as this is noted as an additional indication of reliability (Leger and Thivierge, 1988; Terbizan et al., 2002). Failure to clean data with a transparent system may present skewed data. Between subject numbers reduced from n = 10 to n = 6 for HR and n = 8 for BF at the highest velocities (i.e. ≥ 10 km·h^-1) while in contrast no data was removed in the within subject testing. Though not formally assessed, the disparity between the two testing designs and number of useable data sets warrants further discussion. Increased number of errors for HR and BF variables between subject could have occurred due to the data signal that the monitoring device requires becoming corrupted by varying cross subject movement artefacts (Cho et al., 2011; Witt et al., 2006) such as; EMG activity (Boudet and Chamoux, 2000; McArdle et al., 2009), changes in the mechanics of breathing (McArdle et al., 2009; McCool et al., 2002) or movement of the monitoring device (Clarenbach et al., 2005; Leger and Thivierge, 1988). The full data set and stronger reliability results from the single subject design attained suggests that inter subject differences may influence the device’s ability to collect precise information. Body type was not formally assessed though anecdotally the within subject participant, from which the full data set was attained, possessed ectomorph characteristics. Although firm conclusions cannot be drawn from this issue further work should be completed on the effects of body shape, generic user set-up information and data credibility. Moreover, in this study the participant sex was controlled therefore results are limited and further investigation in to the reliability of the device in other populations should be considered.

Velocity specific analysis allows for identification of micro level limits in the equipment though at times data sets begin to reduce in number and this can affect the statistical analysis. For example, low correlation values within inter and intra reliability velocity specific analysis, if reported out of context, could be misinterpreted though could attributed to low number of data points especially as when data was amalgamated r values were deemed strong.

Some of the variation in the data collected can be attributed to a number of sources and needs to be factored in to any analysis of new technology. Inter and intra subject biological variation (i.e. circadian rhythm, fatigue or subject motivation) and general "noise" from the testing environment (i.e. EMG) can influence reproducibility of data. Additionally some technical error will exist which is outside the control of the researcher all of which influence statistics outcomes and conclusions drawn (Achten and Jeukendrup, 2003; Hopkins et al., 2000; Massin et al., 2000).

The Bioharness^™ can be considered a reliable device within the limitations presented in this study. Within subject reliability data is very strong suggesting the fit of the device on different individuals could be an important factor in attaining consistent data especially for HR and BF. Even considering the latter, some caution regarding data quality should be noted when physical activity is > 10 km·h^-1. This increased variance at higher velocities may have implications for the devices use in sporting contexts though technical upgrades of newer versions of the Bioharness^™ may improve this issue . Being able to access a reliable and valid monitor which measures a range of physiological variables simultaneously in free living conditions will allow for further invaluable understanding of human performance in a variety of environments.

The Bioharness^™ device is designed to enable naturalistic physiologically based monitoring to occur across differing free movement scenarios without the need for obtrusive invasive equipment. The design limitations associated with incorporating multi-variable monitoring within a device which must be unobtrusive to the wearer may place some limitations on the effectiveness of the functioning of individual elements. Free movement physiological data, which the Bioharness^™ aims to capture, is inherently variable (Welk et al., 2004). Therefore, having established the validity and reliability of the Bioharness^™ device in a controlled laboratory environment the authors suggest that the next progression for the Bioharness^™ device is assessment in a less controlled, field based setting. This will allow for a more comprehensive understanding of its capacities in the mode of use it was intended for.