A wearable electrochemical sensor for the real-time measurement of sweat sodium concentration

Abstract

We report a new method for the real-time quantitative analysis of sodium in human sweat, consolidating sweat collection and analysis in a single, integrated, wearable platform. This temporal data opens up new possibilities in the study of human physiology, broadly applicable from assessing high performance athletes to monitoring Cystic Fibrosis (CF) sufferers. Our compact Sodium Sensor Belt (SSB) consists of a sodium selective Ion Selective Electrode (ISE) integrated into a platform that can be interfaced with the human body during exercise. No skin cleaning regime or sweat storage technology is required as the sweat is continually wicked from the skin to a sensing surface and from there to a storage area via a fabric pump. Our results suggest that after an initial equilibration period, a steady-state sodium plateau concentration was reached. Atomic Absorption Spectroscopy (AAS) was used as a reference method, and this has confirmed the accuracy of the new continuous monitoring approach. The steady-state concentrations observed were found to fall within ranges previously found in the literature, which further validates the approach. Daily calibration repeatability (n = 4) was ±3.0% RSD and over a three month period reproducibility was ±12.1% RSD (n = 56). As a further application, we attempted to monitor the sweat of Cystic Fibrosis (CF) sufferers using the same device. We observed high sodium concentrations symptomatic of CF (∼60 mM Na⁺) for two CF patients, with no conclusive results for the remaining patients due to their limited exercising capability, and high viscosity/low volume of sweat produced.

---

Introduction

On a timescale of minutes and hours, the monitoring of sweat electrolyte concentrations can yield much information on the physical and chemical state of the human body. Blood plasma and sweat electrolyte concentrations are related. Low plasma water content (dehydration) and low sodium concentrations (hyponatremia) can result if these are not replaced effectively. These conditions can be detrimental to human health and lead to reduced physical and mental performance. For example, dehydration can manifest itself as an increased sodium concentration in sweat during exercise, so the possibility of monitoring this quantitatively in real-time is an appealing prospect.¹ Currently, for studies involving athletic performance, available fluid volumes are usually inferred only from body weight loss and urine output volumes over a set period of exercise, and electrolyte ion concentrations are only measured separately and retrospectively.^2–5 Measuring ion concentrations directly in bodily fluids such as sweat, inherently gives information on both electrolyte volume and concentrations in plasma, and serves to indicate if these are in order. However, this approach is limited largely due to the unavailability of in situ and/or real-time electrolyte measurement devices and methods. Such devices could be analogous to common heart rate monitors that are widely in use by amateurs and professionals alike. The formulation of specialist sports drinks (e.g. isotonic drinks) and their optimal use in terms of timing and quantity consumed around exercise, ties in with this area of research. Such drinks are generally based on water, carbohydrates and sodium (Na⁺). Carbohydrates are an important energy source for athletes but result in water loss whereas sodium facilitates the uptake of water.^6,7 Sweat monitoring and analysis in a convenient and in situ manner could play a major role in facilitating accurate and timely control of hydration, and therefore enhance performance.

Electrolyte levels in sweat including Na⁺ levels vary considerably depending on the point of the body where they are measured. For example the forehead range was previously found to be as high as 56.7 ± 28.9 mmol l⁻¹ Na⁺ whereas the lower back typically gives half these values.⁸ Additionally, Na⁺ concentration varies widely depending on genetic predisposition, diet and heat acclimatization rate. Conversely, sex and aging do not appear to have a large effect on sweat electrolyte concentrations and no definitive correlation is apparent.^1,9,10 Interestingly, dietary salt intake and sweat rate also do not specifically correlate to sweat electrolyte concentrations.¹¹ These factors underline the individual nature of absolute sweat electrolyte concentrations, and coupled with exercise induced changes over a narrow timescale, make sweat concentrations difficult to interpret, due to the multitude of factors that could be influencing the observed values.

Clinical interest in sweat electrolyte analysis includes the diagnosis of Cystic Fibrosis (CF), indicated by abnormally high sodium levels in sweat.¹² In CF diagnosis, specific sweat electrolyte concentrations (usually Na⁺ or Cl⁻) are determined in a technically complex manner (outlined below) that is not in real-time. Because a specific concentration threshold is available and a small number of analyses are needed for confirming CF in each patient, accuracy is of singular importance – already provided by existing methods of analysis. However, in the area of treatment (e.g. rehabilitative physical exercise for CF patients) and monitoring, high frequency real-time sweat electrolyte data would again be useful.

In order to measure the concentration of specific electrolytes in sweat, there are 3 necessary steps regardless of approach. These are sweat stimulation, sweat collection and/or storage and finally analytical measurement.

Sweat stimulation – Stimulating sweat for subsequent chemical analysis is most commonly performed by methods based on the electrochemical Gibson and Cooke technique from the 1950s.¹² Pilocarpine iontophoresis is based on introducing a cholinergic drug into the skin via an electrical current, thereby stimulating artificial sweating. For non-clinical purposes, exercise like running or cycling generates sweat. Controlled thermal (sauna) methods may also be used.¹³ Each approach has advantages and disadvantages. Sweating is an intrinsic part of exercise therefore obtaining real-time samples during athletic performance does not involve any additional intervention. Conversely, CF sufferers can only exercise for a very short period due to impaired lung function. In addition, CF sufferers tend to generate less sweat, and, because of the elevated electrolyte concentrations, the sweat viscosity is also high, and therefore obtaining valid samples is more difficult. The pilocarpine technique is therefore most often used clinically. However, specialist equipment and training is required and in some instances discomfort ranging from skin reddening to burns have been reported.¹⁴

Sweat collection and/or storage – This stage has by far the greatest impact on the accuracy of electrolyte analysis. Sweating helps regulate body temperature via the evaporation of sweat from the skin surface. It is this evaporation that can easily lead to inaccurate high electrolyte concentration readings and is the single biggest potential source of error in this field of research. Electrolyte concentration also varies greatly with the sampling site on the body, so once chosen, previous salt deposits must be thoroughly cleaned from the sampling site, and the area isolated from run-off sweat that has originated from other areas e.g. Sweat originating from the head, which passes down the back.⁸ Once sweating has started, an absorbent material is sealed against the chosen sampling area with an impermeable plastic backing to prevent vapour loses. The absorbent material may be a gauze or cellulosic material like in the sweat patch PharmChemTM system.¹⁵ Such patches either swab sweat briefly or can be left in place to accumulate sweat over hours. The sweat collected is carefully weighed, extracted and diluted for analysis. Volume losses with paper and gauze methods can be as high as 2% per minute.¹¹ Alternatively a container with a known amount of water is sealed against the skin and sweat is accumulated in solution over a known time. Sweat may also be collected from greater portions of the body by plastic sleeves covering an entire arm or armpit. Knowledge of whole body sweat electrolyte concentrations is a desirable measurement parameter but sample collection can be very inconvenient.¹⁶ In one approach, a box lined with plastic contains an exercise bicycle. All equipment and the subject are carefully washed before exercise. Following exercise, the box and equipment, clothes and the subject are again washed. In this way all electrolytes produced on the body are collected for analysis. However, this only gives one value, which is the total amount of electrolyte released over the entire exercise period. The Macroduct system goes some way towards eliminating evaporation losses and contamination by surrounding dried sweat.¹⁷ In a previous study on CF conducted by our research group, iontophoresis on the wrist was followed by applying the macroduct system.¹⁸ This consists of a coiled length of capillary tubing mounted on a wrist strap. A small orifice on the underside is held in place against the skin and sweat enters and mixes within the coil. Here it may stay until required for analysis with minimal evaporation losses. Due to the small collection orifice, contamination is minimised. One particular drawback of this and the other collection techniques mentioned above, as relevant to this paper, is that temporal data cannot be obtained conveniently, unless sensors are incorporated into the sampling unit. The intimate combination of sampling and analysis of sweat was previously attempted by our group albeit using separate devices in the first instance.¹⁸

Analytical measurement – Stored sweat samples are usually extracted and diluted quantitatively prior to analysis (indirect analysis). Sodium in sweat is mostly analysed by Flame Photometry techniques.¹¹ Another major technique used for indirect sodium analysis is based on Ion Selective Electrodes (ISEs).^19,20 Our own group measured sodium amongst other parameters in sweat using the Macroduct collector and a solid-state ISE array and reference electrode.¹⁸ Although the results were promising, sweat was measured after collecting a limited volume, meaning that replicate and temporal data was not available. Others have focussed on sodium's main counter ion in sweat, namely chloride using ISEs.^21–23 These methods typically employed iontophoresis for sweat stimulation with the analysis taking place at a later time. Invariably, evaporation was cited as the main error source regarding accuracy. Regarding precision, the availability of insufficient sweat volume caused the most variability and noise in the analytical signal, despite allowing a reasonable period for sweat accumulation. To date the US Cystic Fibrosis Foundation (CFF) has not approved ISEs for diagnosing CF, presumably due to the difficulties mentioned above. More recently, sodium quantification in sweat was carried out using Capillary Zone Electrophoresis (CZE).^15,24 Additional techniques for measuring sweat ions include titration based methods such as amperometric-coulometric determination of chloride using a chloridometer.²⁵ Conductivity and osmolarity measurements have also been used.¹¹ Both are largely confined to screening in clinical applications and are not approved by the CFF. Wescor's sweat conductivity analyzer is one such technique of note.²⁶ Sweat was induced by 60 min of exercise and collected by closed-pouch collector. The analysis reflects total ionic content given in “NaCl equivalent” units (non-standard units). Due to the non-selective nature of these techniques, there is often positive systematic error. Although useful as a simple clinical screening method (for CF), it cannot yield quantitative analyte specific data in real-time. To the best of our knowledge the only on-skin electrochemical probes for specific electrolytes commercially available are those for measuring pH such as the skincheck1TM (http://www.hannainst.co.uk). Our own research group has achieved some success with real-time monitoring of sweat pH in a wireless fashion.²⁷ In a similar study an attempt was made to monitor sodium in sweat wirelessly. Although compact solid-state technology was used, the sodium data (ISE based) exhibited non-Nernstian slopes, severe drift and poor reproducibility. This is attributed to the fact that solid-state ISEs in the form used were inherently less stable (both working and reference electrodes) than their classic inner solution based counterparts. Our research in this field continues.^28–31

In this paper, we present our ISE based Sweat Sensor Belt (SSB) with which we aim to consolidate sweat collection and analysis conveniently in a single device and for the first time access reliable real-time electrolyte data during exercise.

Experimental

Sweat sensor belt (SSB)

The Sodium Sensor Belt (SSB) consists of an impermeable plastic platform holding the ISEs and sweat wicking materials (pump), together with a potentiometer. The assembly was held against the subject's lower back during analysis by elasticated belt (Fig. 1).

Fig. 1 The Sodium Sensor Belt (SSB) showing (from top to bottom) a schematic of the impermeable holder platform with lower orifice, the entire sensor assembly showing signal display and the SSB mounted onto the lower back.

The platform consists of impact grade acrylonitrile butadiene styrene (ABS) plastic platform fabricated via a prototyping 3D printer. This light, rigid yet rather porous plastic was rendered impermeable by immersion in dimethylketone for 15 min. The fluid handling system is based on the use of fabrics with inherent moisture wicking properties through capillary action. A patch of polyimide/lycra® blend (http://www.sofileta.com), common in sportswear (from now referred to as just lycra), was glued to the reverse side of the platform for contact with human skin. This wicks fresh sweat produced through an orifice in the platform towards the sensor tips and finally into a terminal waste reservoir as shown in Fig. 1. The reservoir material contains beads of Absorptex (http://www.smartex.ie) with a free swell capacity of 25 g g⁻¹. A waste patch of 15 × 30 mm was sufficient to absorb sweat/water for 3 h at typical sweat rates. The platform serves to press the patch against the skin to capture sweat efficiently as it emerges from skin pores and through the rim forming a seal against the skin, ensuring that only sweat from beneath the area of the platform is measured. This is necessary to prevent mixing of sweat on the back caused by sweat beads running down the subjects back. This arrangement also minimised evaporation losses prior to sensing.

The ISE electrodes were fabricated in-house using PVC tubing as barrels. The reference electrode was filled with 0.1 M KCl on the day of analysis and the tip plugged with porous compressed foam or vycor glass supplied by Biomedical Systems (code MF-2064). Electrodes were stored in appropriate filling solutions filling solution when not in use.

All chemicals were supplied by Sigma and were analytical grade. The electrodes were connected to a potentiometer (Tenma 72-7935) supplied by Farnell Ireland Ltd.

Potentiometric membranes were prepared using 250 mg 2-Nitrophenyl octyl ether, 125 mg PVC, 6.5 mmol kg⁻¹ 4-tert-Butylcalix[4]arenetetraacetic acid tetraethyl ester (Sigma 420484) and 2.7 mmol Kg⁻¹ potassium tetrakis(4-chlorophenyl) borate dissolved in dry THF and evaporated slowly. The well known calixarene ionophore was developed by members of our research group.³² The electrochemical cell had the following arrangement:

Ag|AgCl|0.1M KCl‖sampling unit|PVC membrane|0.1 M NaCl|AgCl|Ag. Membranes were conditioned in 0.1 M sodium chloride for 12 h and deionised water for half an hour prior to use. Two calibrations before and two after each trial (total 4 calibrations) were carried out using standard solutions. This also was used to check that the sensor was functioning properly. The Calibration range was based on standards 20, 40, 60, 80 and 100 mM NaCl which covers the typical range of sodium concentration in sweat. Inter-conversion between sodium activity and concentration values was carried out according to Debye–Hückel equations.

Cycle exercise trials

The SSB was strapped to the lower back after calibration. Time and potential values were noted periodically during trials. The lycra sweat pump was dampened a little with de-ionised water to keep the ISE tips moist at the start of each trial and to give reasonably stable starting potentials. A standard exercise bicycle was used. For the convenience and safety of the subjects who volunteered for the trials, the pace and duration of each trial was self-selected. Healthy subjects typically cycled for 1 h. Sufferers of Cystic Fibrosis (CF) cycled for considerably less time. Medical staff were present for all trials involving CF patients.

AAS method

Pre-weighed 1 cm² patches of lycra were soaked in either NaCl calibration standards (same solutions as used for the ISE calibration) or used to swab sweat samples from the lower back. The volume of liquid absorbed was estimated by re-weighing and assuming a density of 1.0 g ml⁻¹. Extraction followed and involved shaking each patch twice in 20 ml de-ionised water for 15 s. Following this, 10 ml water was added to the combined extract. All solutions were analysed in triplicate according to the manufacturers' manual guidelines (Varian). A wavelength of 589.6 nm was used.

Results and discussion

The use of Ion Selective Electrodes (ISEs) is ideal for measuring cation and anion concentrations in aqueous based samples including sweat. Our group has much experience with ISEs, from developing new ionophores to their applications in environmental, industrial and physiological measurements.^18,32–38 One of the most successful of these ionophores is a calix[4]arene based sodium ionophore, which we use in the SSB.³²

In our original design of the sensor ISE, we realised that in conjunction with the PVC based working electrode, the reference electrode tip should provide efficient ionic conductivity between the inner filling solution and the passing sweat on the lycra patch with which the sensor makes contact. We initially used a compacted polyurethane foam plug to serve this function. The sensor response was excellent, showing Nernstian calibration slopes. However the bleed rate of inner filling solution when in contact with the lycra sweat pump was unacceptably high with all filling solution drained before a single exercise trial could be completed. The polyurethane plugs were therefore replaced with vycor glass tips. Over a 3 month period of trials, no discernible liquid loss occurred from the reference electrode. The resulting calibration slopes were sub-Nernstian but constant at around 50 mV per decade of sodium activity. Even after a 3 month period of sensor trials, Nernstian calibration slopes could be obtained when the vycor tips were exchanged for foam, demonstrating that the ISE still worked. Fig. 2A shows an example of the calibration repeatability typically achieved, based on the four calibrations for a single trial.

Fig. 2 Precision in terms of repeat calibrations. A. Typical repeatability (n = 4, 1 day). B. Reproducibility (n = 56, 98 day period). a = activity.

To represent the typical daily precision achievable, we take an intermediate sodium standard (40 mM, or more accurately a Debye–Hückel based activity of 0.033) representing typical sweat sodium levels. For this standard, a mean potential of −10.8 ± 0.8 mV was obtained (n = 4) as shown in Fig. 2A. Given the slope of the line, this translates into a repeatability of 0.033 ± 0.001 in terms of activity or a % relative standard deviation (% RSD) of 3.0%. The use of activity is appropriate in this instance as activity coefficients are considerably below unity when approaching a sodium chloride concentration of 0.1 M. In terms of reproducibility over a 98 day period and a total of 56 calibrations, the mean absolute voltages and precision error bars are shown in Fig. 2B. In this case, the sodium standard of activity 0.033 gave a mean potential of −11.1 ± 2.7 mV. This translates into a reproducibility of 0.033 ± 0.004 or a % RSD of 12.1% (n = 56). Clearly, the frequency and rigour of calibration must be chosen according to the required precision and accuracy of any eventual sensor application (e.g. Obtaining absolute concentrations is analytically more demanding than monitoring concentration changes). Good linearity was achieved, R² exceeding 0.99 in all cases. A darkening of the vycor reference tips was observed over time possibly due to bacterial growth. Contact with the sweat matrix possibly contributed to this. An antibacterial agent could be added to the storage solutions to prevent this, if necessary.

Atomic Absorption Spectroscopy (AAS) was chosen as a reference and validation method to assess the performance of our SSB. Pre-weighed 1 cm² patches of lycra were soaked in each of the NaCl calibration standards and re-weighed.

Following an extraction and dilution procedure, the samples were analysed in triplicate by calibrated AAS to give the results shown in Table 1.

Table 1 AAS % recovery (accuracy) values as determined by the analysis of lycra patches soaked in the specified NaCl standard

Standard/mM	Observed [Na]/mM	% Recovery
20.0	20.9	104.5
40.0	40.5	101.3
60.0	58.1	96.8
80.0	81.6	102.0
100.0	103.7	103.7

---

It can be seen that good accuracy is achieved with all results within a 95–105% accuracy specification. The same procedure was employed using real sweat samples swabbed directly onto lycra patches from exposed skin at intervals during a cycling trial. Samples were taken from the immediate vicinity of our belt mounted sensor on the lower back and immediately sealed into plastic containers to minimise evaporation and other sample perturbation. Weighing, extraction and AAS analysis took place as soon as possible after a trial. Just prior to sample weighing, the same lycra patches were analysed on the bench by direct contact with the ISEs which were temporarily removed from the SSB. This served to directly compare our electrochemical sensor performance to the AAS reference measurements using real-life samples. The results of this exercise are shown in Table 2. It is apparent that the ISE and AAS results obtained are similar, although considerable scatter is apparent. This may be due to some drainage of sweat from patches during storage, meaning that low sample volume may have affected ISE contact and signal quality. However, a paired t-test suggests there is no significant difference between the two methods as the calculated t value is 0.301 and the probability of this result, assuming the null hypothesis, is 0.771 (n − 1 = 8 degrees of freedom).

Table 2 Sodium concentrations measured from sweat swabs by both AAS and the ISE from the SSB during an exercise trial

Trial time (minutes)	ISE [Na]/mM	AAS [Na]/mM
18	56.8	71.8
26	68.2	66.7
31	70.5	66.3
37	61.5	70.9
43	82.6	73.1
47	60.3	63.2
50	59.4	58.5
52	83.8	72.6
58	76.3	68.4
Mean	68.8	67.9

---

A total of 4 healthy and 4 subjects with CF took part in trials (all male). Useful data was only obtained from trials where the subject produced enough sweat, based on individual physiology and willingness to exercise with sufficient vigour for periods exceeding 30 min or so. The SSB was strapped to the subject's lower back and cell potential values were noted from a digital read-out beside the sensor platform as shown in Fig. 1. Typically, following the visual appearance of sweat beads on the subjects back, the potential rose until a concentration plateau was reached about 15 min after this. Fig. 3 shows examples of sodium concentration change with time during cycling exercise trials.

Fig. 3 Cycle exercise trials showing rising sodium concentration with time. A concentration plateau is reached as indicated. Example of A. Healthy subject trial and B. CF subject trial.

For all full trials, the plateau sodium concentrations changed very little for the remaining duration of the trials. The constant nature of the sodium concentration at this stage suggested that the initial small quantity of de-ionised water (dampened lycra) was no longer causing a dilution effect. Secondly the plateau suggests that significant evaporation or sodium accumulation was not occurring and fresh sweat was being analysed as desired. This is bolstered by the fact that for all trials, concentrations fell within the range of previous literature values for the lower back. Nimmo et al. state a mean of 26.2 ± 19.4 mM for lower back sodium concentration for cycling males.⁸ Sawka et al. suggest 35 mM as a universal mean for human sweat sodium.¹Table 3 summarises the sodium plateau concentrations observed for all trials. The high SSB sodium concentrations of around 60 mM as observed for the CF subjects in Table 3 is in agreement with the literature, whereby concentrations around a diagnostic threshold of 60 mM are regarded as CF positive.¹² The quality of the SSB data is best for healthy subjects probably due to a higher number of plateau data points as longer trial times are possible. Only 2 of the 4 subjects with CF managed to complete a successful trial, and this trial is considerably shorter than for healthy subjects. The nature of CF prevents sufferers from performing physical exercise of the same duration and intensity as healthy individuals due to mucus build-up in the lungs which impairs intensive physical activity. Despite this, regular exercise, particularly in young CF sufferers, is important as it improves overall lung function and quality of life.

Table 3 Cycle exercise trial results showing SSB plateau sodium concentrations. Proximal sweat swabs were analysed by AAS in some cases^a

Subject	Trial end time/min	Health	SSB Plateau [Na^+]/mM	AAS swab [Na^+]/mM
a N = healthy male CF = male with Cystic Fybrosis AAS swabs taken on bare skin beside SSB. b Based on AAS values from Table 2.
1	55	N	25.4 ± 0.6	—
2	58	N	32.5 ± 1.5	68 ± 5
3	60	N	18.0 ± 1.2	—
4	60	N	24.8 ± 1.2	72 ± 5
5	36	CF	54.5 ± 4.4	—
6	40	CF	61.9 ± 12.5	284 ± 71

---

For all methods of sweat electrolyte analysis, the sweat collection and storage step is perhaps the greatest single source of error as sweat bead movement, regional intermixing and rapid evaporation cause large fluctuations in sample concentrations: For example, a 2.5 h marathon runner of 70 kg body weight, secretes about 2 litres of sweat per hour, with ca. 80% of the water evaporating (cooling function) and 20% dripping away (wasted).³⁹ Evaporation is therefore a serious issue, as it can make correlation with changes in physiological fluids like plasma difficult to achieve. Our strategy of capturing and analysing sweat composition in a defined location, with minimal external contaminations, and in real-time therefore has the potential to overcome these artefacts and provide accurate information on electrolyte loss in sweat. The results also suggest that the sweat pump appears to be effective at delivering fresh sweat sample past the sensor ISE tips and on to the absorptex waste compartment over the timescale of the exercise.

The AAS reference method, through the analysis of sweat swabbed directly from the bare skin demonstrates how vital a robust sampling strategy is. Although the AAS component of the analytical method was shown to be accurate (Table 1), the associated sampling approach was wholly inadequate. The quality of the AAS results was in sharp contrast to those obtained using the SSB. The AAS sodium concentrations in Table 3 are all much higher than the corresponding SSB values and unlike the SSB values, are outside reasonable ranges as suggested by the literature. These high values are most likely due to rapid evaporation of sweat leading to sodium accumulation on the skin. The poor precision of the AAS samples (Table 3) further reflects the ever-changing concentrations and suggest mixing of sweat from different body regions (dripping and rolling sweat) combined with the evaporation effects. The great variety and ongoing search of new sweat processing methods in the literature attest to the importance of a good sampling approach.

The reversibility of the SSB was investigated to determine whether sample carryover would be an issue in the patch (i.e. would a higher concentration affect the signal obtained from a subsequent return to a lower concentration). The SSB was mounted in the usual upright position on a waterproof surface. Sodium standards of varying concentration were delivered to the absorbent patch on the reverse side of the sensor by peristaltic pump. The flow rate was 17 μl min⁻¹ to approximately mimic the rate of excretion of sweat given the surface contact area of the sensor.¹ The standard supply line was only removed from the SSB for purging at each concentration change. The experiment was conducted over a 3 h period, repeatedly cycling through the entire concentration range from 20–100 mM NaCl. Fig. 4 shows the resultant plot. Even after a 3 h period of continuous sweat pumping, good reversibility was observed with little evidence of hysteresis, sodium accumulation or evaporation within the analysis timeframe. The potential typically changes no more than ±1 mV for repeat analysis of any concentration. The Absorptex waste terminal demonstrated considerable swelling suggesting an effective sweat wicking action. The range of concentration change and trial duration is lower in real-life trials proving the sensor has adequate reversibility.

Fig. 4 The reversibility of the SSB is proven over a 3 h fluid collection period.

We also wished to find out whether the pressure with which the ISE tips were pushed against a sample surface made a difference to the analytical measurement. This matter is related to whether one can freely interchange between analysing a bulk solution and a solution soaked fabric patch. Previously, Bray et al. reported that for volumes greater than 1 μl between skin and sensor, the pressure with which the electrode was pushed to the sample had no significant impact on the potential output. It was also revealed that authors who initially stated that the applied pressure was critical when using ISEs, demonstrated that they had insufficient sweat volumes available for reliable readings.²² Insufficient liquid would impair thorough sample contact with the sensor tips, hence introducing error. In using our ISE based sensor, we found that within a reasonable operational range, pressure made no difference to the analytical measurements. Consequently, our ISEs yielded the same absolute potential values for calibration standards whether placed in bulk solution or held against standard soaked lycra fabric.

In general, potential values observed were very stable, no doubt thanks in part to the high sodium activities characteristic of sweat analysis (>0.01), at which all ISEs function very well.

What most analytical methods for sweat analysis have in common is that there is a considerable time delay between the emergence of sweat from the skin and its analysis. For any such approach, evaporation and other sample perturbation will always remain a major source of analytical error. The strategy of minimising the time delay between sweat emergence on the skin surface and analysis – analysing fresh sweat, inherently reduces error.

For clinical conditions, concentration thresholds and ranges are better defined than for healthy individuals. For example as relevant to our current study, sodium concentrations at or above 60 mM Na⁺ are indicative of CF disease.¹² It is only possible to assign an absolute value here because of the striking difference to ‘normal’ Na⁺ levels in humans – it is approximately double typical healthy values. Interestingly, unlike in CF diagnosis, the literature contains few ‘recommended’ or ‘healthy’ absolute electrolyte concentrations in the sweat of healthy individuals because, as stated before, if using electrolytes in sweat to assess the state of other physiological fluids, the greatest variation occurs depending on the location on the body where sweat is collected in addition to varying greatly from individual to individual for reasons that are not fully understood.⁸ The former variable is dealt with to an extent by carefully sampling sweat from a specific area of the body (e.g. lower back). The issue of individuality clearly persists as shown by our study as well as others. Nimmo et al. measured sodium concentration on the lower back of 10 healthy males performing a similar cycling exercise routine as our subjects.⁸ Results indicated sodium concentrations in the range 26.2 ± 19.4 mM, a variation of ±74%. The best strategy therefore may be to determine baseline or ‘healthy’ electrolyte levels for each individual. In terms of the SSB, this may correspond to the individual sodium plateau concentration. It is then the deviations from the individual plateau concentrations over time, which may serve for example as an indicator of hydration level. It is only through gathering a body of accurate and precise data on large numbers of individuals that we can begin to gain a better understanding of this aspect of physiology.

Conclusion

The real-time analysis of sweat electrolytes viain situ measurements using an on-body sampling and sensing platform has been demonstrated, and may open up new applications in research areas such as athletic performance and healthcare. Exercise trials using the platform have generated values of sodium concentrations in sweat, in agreement with literature ranges and verified using an AAS reference method. In future we would like to use the on-body sensing system to monitor the onset of dehydration, and for this exercise trials would have to be extended with appropriate health safeguards in place.

A large number of cations and anions can be selectively analysed according to the many ionophores available for use in ISEs. These could simply replace our current sodium selective ionophore or be used as part of a sensor array. The miniaturisation and placement of sensor tips in closer proximity to the skin may reduce the delay time between excretion and analysis of sweat. In addition to inherently reducing sample evaporation and perturbation, it may render our sensor better for clinical diagnosis and monitoring such as in the case of CF, where only relatively short periods of exercise can be expected of patients and correspondingly lower sweat volumes are available for analysis. In parallel, the integration of wireless communications is relatively straightforward and could result in truly individualised applications.

Acknowledgements

We would like to thank Science Foundation Ireland (SFI) for the support and funding to carry out this research. The work was carried out under “CLARITY – Centre for Sensor Web Technologies” – code 07/CE/I1147.

Notes and references

1. N. Sawka, L. M. Burke, E. R. Eichner, R. J. Maughan, S. J. Montain and N. S. Stachenfeld, Med. Sci. Sports Exercise, 2007, 39, 377.
2. R. J. Maughan and S. M. Shirreffs, Science & Sports, 2004, 19, 234.
3. T. Yoshida, T. Takanishi, S. Nakai, A. Yorimoto and T. Morimoto, Eur. J. Appl. Physiol., 2002, 87, 529.
4. B. Sanders, T. D. Noakes and S. C. Dennis, Eur. J. Appl. Physiol., 1999, 80, 318.
5. R. J. Maughan and S. M. Shirreffs, Journal of Sports Sciences, 1997, 15, 297.
6. S. P. von Duvillard, W. A. Braun, M. Markofski, R. Beneke and R. Leithauser, Nutrition, 2004, 20, 651.
7. F. Brouns, W. Saris and H. Schneider, International Journal of Sport Nutrition, 1992, 2, 229.
8. M. J. Patterson, S. D. R. Galloway and M. A. Nimmo, Exp. Physiol., 2000, 85, 869.
9. F. Meyer, O. Baror, D. Macdougall and G. J. F. Heigenhauser, Medicine and Science in Sports and Exercise, 1992, 24, 776.
10. T. Morimoto, Z. Slabocho, R. K. Naman and F. Sargent, Journal of Applied Physiology, 1967, 22, 526.
11. M. Constantinescu and B. C. Hilman, Laboratory Medicine, 1996, 27, 472.
12. L. E. Gibson and R. E. Cooke, Pediatrics, 1959, 23, 545.
13. T. T. Amatruda and L. G. Welt, Journal of Applied Physiology, 1953, 5, 759.
14. J. M. Kirk, Arch. Dis. Child., 2000, 82, 425.
15. B. M. R. Appenzeller, C. Schummer, S. B. Rodrigues and R. Wennig, J. Chromatogr., B: Anal. Technol. Biomed. Life Sci., 2007, 852, 333.
16. S. M. Shirreffs and R. J. Maughan, Journal of Applied Physiology, 1997, 82, 336.
17. M. E. Miller, J. M. Cosgriff and R. H. Schwartz, Clinical Chemistry, 1985, 31, 1715.
18. A. Lynch, D. Diamond and M. Leader, Analyst, 2000, 125, 2264.
19. H. M. Barbour, Annals of Clinical Biochemistry, 1991, 28, 150.
20. H. Northall and G. A. York, British Journal of Biomedical Science, 1995, 52, 68.
21. L. Szabo, M. A. Kenny and W. Lee, Clinical Chemistry, 1973, 19, 727.
22. P. T. Bray, G. C. F. Clark, G. J. Moody and J. D. R. Thomas, Clin. Chim. Acta, 1977, 80, 333.
23. P. T. Bray, G. C. F. Clark, G. J. Moody and J. D. R. Thomas, Clin. Chim. Acta, 1977, 77, 69.
24. T. Hirokawa, H. Okamoto, Y. Gosyo, T. Tsuda and A. R. Timerbaev, Anal. Chim. Acta, 2007, 581, 83.
25. V. A. Legrys and R. W. Burnett, Archives of Pathology & Laboratory Medicine, 1994, 118, 865.
26. P. Boisvert and V. Candas, Eur. J. Appl. Physiol., 1994, 69, 176.
27. D. Morris, S. Coyle, Y. Z. Wu, K. T. Lau, G. Wallace and D. Diamond, Sens. Actuators, B, 2009, 139, 231.
28. D. Desmond, B. Lane, J. Alderman, J. D. Glennon, D. Diamond and D. W. M. Arrigan, Sens. Actuators, B, 1997, 44, 389.
29. C. M. McGraw, T. Radu, A. Radu and D. Diamond, Electroanalysis, 2008, 20, 340.
30. K. Y. Chumbimuni-Torres, N. Rubinova, A. Radu, L. T. Kubota and E. Bakker, Anal. Chem., 2006, 78, 1318.
31. J. Sutter, A. Radu, S. Peper, E. Bakker and E. Pretsch, Anal. Chim. Acta, 2004, 523, 53.
32. D. Diamond, G. Svehla, E. M. Seward and M. A. McKervey, Anal. Chim. Acta, 1988, 204, 223.
33. A. M. Cadogan, D. Diamond, M. R. Smyth, M. Deasy, M. A. McKervey and S. J. Harris, Analyst, 1989, 114, 1551.
34. D. Diamond and M. A. McKervey, Chem. Soc. Rev., 1996, 25, 15.
35. D. Diamond and K. Nolan, Anal. Chem., 2001, 73, 22A.
36. K. M. O’Connor, D. W. M. Arrigan and G. Svehla, Electroanalysis, 1995, 7, 205.
37. B. Schazmann and D. Diamond, New J. Chem., 2007, 31, 587.
38. B. Schazmann, S. O'Malley, K. Nolan and D. Diamond, Supramol. Chem., 2006, 18, 515.
39. R. J. Maughan, Int. J. Sports Med., 1992, 13, S132.

---